Posts Tagged 'Calculating Kernels'

C# How to: Fuzzy Blur Filter

Article Purpose

This article serves to illustrate the concepts involved in implementing a Fuzzy Blur Filter. This filter results in rendering  non-photo realistic images which express a certain artistic effect.

Frog: Filter Size 19×19

Frog: Filter Size 19x19

Sample Source Code

This article is accompanied by a sample source code Visual Studio project which is available for download here.

Using the Sample Application

The sample source code accompanying this article includes a based test application. The concepts explored throughout this article can be replicated/tested using the sample application.

When executing the sample application the user interface exposes a number of configurable options:

  • Loading and Saving Images – Users are able to load source/input from the local system by clicking the Load Image button. Clicking the Save Image button allow users to save filter result .
  • Filter Size – The specified filter size affects the filter intensity. Smaller filter sizes result in less blurry being rendered, whereas larger filter sizes result in more blurry being rendered.
  • Edge Factors – The contrast of fuzzy expressed in resulting depend on the specified edge factor values. Values less than one result in detected being darkened and values greater than one result in detected image edges being lightened.

The following image is a screenshot of the Fuzzy Blur Filter sample application in action:

Fuzzy Blur Filter Sample Application

Frog: Filter Size 9×9

Frog: Filter Size 9x9

Fuzzy Blur Overview

The Fuzzy Blur Filter relies on the interference of when performing in order to create a fuzzy effect. In addition results from performing a .

The steps involved in performing a Fuzzy Blur Filter can be described as follows:

  1. Edge Detection and Enhancement – Using the first edge factor specified enhance by performing Boolean Edge detection. Being sensitive to , a fair amount of detected will actually be in addition to actual .
  2. Mean Filter Blur – Using the edge enhanced created in the previous step perform a blur. The enhanced edges will be blurred since a does not have edge preservation properties. The size of the implemented depends on a user specified value.
  3. Edge Detection and Enhancement –  Using the blurred created in the previous step once again perform Boolean Edge detection, enhancing detected edges according to the second edge factor specified.

Frog: Filter Size 9×9

Frog: Filter Size 9x9

Mean Filter

A Blur, also known as a , can be performed through . The size of the / implemented when preforming will be determined through user input.

Every / element should be set to one. The resulting value should be multiplied by a factor value equating to one divided by the / size. As an example, a / size of 3×3 can be expressed as follows:

Mean Kernel

An alternative expression can also be:

Mean Kernel

Frog: Filter Size 9×9

Frog: Filter Size 9x9

Boolean Edge Detection without a local threshold

When performing Boolean Edge Detection a local threshold should be implemented in order to exclude . In this article we rely on the interference of in order to render a fuzzy effect. By not implementing a local threshold when performing Boolean Edge detection the sample source code ensures sufficient interference from .

The steps involved in performing Boolean Edge Detection without a local threshold can be described as follows:

  1. Calculate Neighbourhood Mean – Iterate each forming part of the source/input . Using a 3×3 size calculate the mean value of the neighbourhood surrounding the currently being iterated.
  2. Create Mean comparison Matrix – Once again using a 3×3 size compare each neighbourhood to the newly calculated mean value. Create a temporary 3×3 size , each element’s value should be the result of mean comparison. Should the value expressed by a neighbourhood exceed the mean value the corresponding temporary element should be set to one. When the calculated mean value exceeds the value of a neighbourhood the corresponding temporary  element should be set to zero.
  3. Compare Edge Masks – Using sixteen predefined edge masks compare the temporary created in the previous step to each edge mask. If the temporary matches one of the predefined edge masks multiply the specified factor to the currently being iterated.

Note: A detailed article on Boolean Edge detection implementing a local threshold can be found here:

Frog: Filter Size 9×9

Frog: Filter Size 9x9

The sixteen predefined edge masks each represent an in a different direction. The predefined edge masks can be expressed as:

Boolean Edge Masks

Frog: Filter Size 13×13

Frog: Filter Size 13x13

Implementing a Mean Filter

The sample source code defines the MeanFilter method, an targeting the class. The definition listed as follows:

private static Bitmap MeanFilter(this Bitmap sourceBitmap, 
                                 int meanSize)
{
    byte[] pixelBuffer = sourceBitmap.GetByteArray(); 
    byte[] resultBuffer = new byte[pixelBuffer.Length];

double blue = 0.0, green = 0.0, red = 0.0; double factor = 1.0 / (meanSize * meanSize);
int imageStride = sourceBitmap.Width * 4; int filterOffset = meanSize / 2; int calcOffset = 0, filterY = 0, filterX = 0;
for (int k = 0; k + 4 < pixelBuffer.Length; k += 4) { blue = 0; green = 0; red = 0; filterY = -filterOffset; filterX = -filterOffset;
while (filterY <= filterOffset) { calcOffset = k + (filterX * 4) + (filterY * imageStride);
calcOffset = (calcOffset < 0 ? 0 : (calcOffset >= pixelBuffer.Length - 2 ? pixelBuffer.Length - 3 : calcOffset));
blue += pixelBuffer[calcOffset]; green += pixelBuffer[calcOffset + 1]; red += pixelBuffer[calcOffset + 2];
filterX++;
if (filterX > filterOffset) { filterX = -filterOffset; filterY++; } }
resultBuffer[k] = ClipByte(factor * blue); resultBuffer[k + 1] = ClipByte(factor * green); resultBuffer[k + 2] = ClipByte(factor * red); resultBuffer[k + 3] = 255; }
return resultBuffer.GetImage(sourceBitmap.Width, sourceBitmap.Height); }

Frog: Filter Size 19×19

Frog: Filter Size 19x19

Implementing Boolean Edge Detection

Boolean Edge detection is performed in the sample source code through the implementation of the BooleanEdgeDetectionFilter method. This method has been defined as an targeting the class.

The following code snippet provides the definition of the BooleanEdgeDetectionFilter :

public static Bitmap BooleanEdgeDetectionFilter( 
       this Bitmap sourceBitmap, float edgeFactor) 
{
    byte[] pixelBuffer = sourceBitmap.GetByteArray(); 
    byte[] resultBuffer = new byte[pixelBuffer.Length]; 
    Buffer.BlockCopy(pixelBuffer, 0, resultBuffer, 
                     0, pixelBuffer.Length); 

List<string> edgeMasks = GetBooleanEdgeMasks(); int imageStride = sourceBitmap.Width * 4; int matrixMean = 0, pixelTotal = 0; int filterY = 0, filterX = 0, calcOffset = 0; string matrixPatern = String.Empty;
for (int k = 0; k + 4 < pixelBuffer.Length; k += 4) { matrixPatern = String.Empty; matrixMean = 0; pixelTotal = 0; filterY = -1; filterX = -1;
while (filterY < 2) { calcOffset = k + (filterX * 4) + (filterY * imageStride);
calcOffset = (calcOffset < 0 ? 0 : (calcOffset >= pixelBuffer.Length - 2 ? pixelBuffer.Length - 3 : calcOffset)); matrixMean += pixelBuffer[calcOffset]; matrixMean += pixelBuffer[calcOffset + 1]; matrixMean += pixelBuffer[calcOffset + 2];
filterX += 1;
if (filterX > 1) { filterX = -1; filterY += 1; } }
matrixMean = matrixMean / 9; filterY = -1; filterX = -1;
while (filterY < 2) { calcOffset = k + (filterX * 4) + (filterY * imageStride);
calcOffset = (calcOffset < 0 ? 0 : (calcOffset >= pixelBuffer.Length - 2 ? pixelBuffer.Length - 3 : calcOffset));
pixelTotal = pixelBuffer[calcOffset]; pixelTotal += pixelBuffer[calcOffset + 1]; pixelTotal += pixelBuffer[calcOffset + 2]; matrixPatern += (pixelTotal > matrixMean ? "1" : "0"); filterX += 1;
if (filterX > 1) { filterX = -1; filterY += 1; } }
if (edgeMasks.Contains(matrixPatern)) { resultBuffer[k] = ClipByte(resultBuffer[k] * edgeFactor);
resultBuffer[k + 1] = ClipByte(resultBuffer[k + 1] * edgeFactor);
resultBuffer[k + 2] = ClipByte(resultBuffer[k + 2] * edgeFactor); } }
return resultBuffer.GetImage(sourceBitmap.Width, sourceBitmap.Height); }

Frog: Filter Size 13×13

Frog: Filter Size 13x13

The predefined edge masks implemented in mean comparison have been wrapped by the GetBooleanEdgeMasks method. The definition as follows:

public static List<string> GetBooleanEdgeMasks() 
{
    List<string> edgeMasks = new List<string>(); 

edgeMasks.Add("011011011"); edgeMasks.Add("000111111"); edgeMasks.Add("110110110"); edgeMasks.Add("111111000"); edgeMasks.Add("011011001"); edgeMasks.Add("100110110"); edgeMasks.Add("111011000"); edgeMasks.Add("111110000"); edgeMasks.Add("111011001"); edgeMasks.Add("100110111"); edgeMasks.Add("001011111"); edgeMasks.Add("111110100"); edgeMasks.Add("000011111"); edgeMasks.Add("000110111"); edgeMasks.Add("001011011"); edgeMasks.Add("110110100");
return edgeMasks; }

Frog: Filter Size 19×19

Frog: Filter Size 19x19

Implementing a Fuzzy Blur Filter

The FuzzyEdgeBlurFilter method serves as the implementation of a Fuzzy Blur Filter. As discussed earlier a Fuzzy Blur Filter involves enhancing through Boolean Edge detection, performing a blur and then once again performing Boolean Edge detection. This method has been defined as an extension method targeting the class.

The following code snippet provides the definition of the FuzzyEdgeBlurFilter method:

public static Bitmap FuzzyEdgeBlurFilter(this Bitmap sourceBitmap,  
                                         int filterSize,  
                                         float edgeFactor1,  
                                         float edgeFactor2) 
{
    return  
    sourceBitmap.BooleanEdgeDetectionFilter(edgeFactor1). 
    MeanFilter(filterSize).BooleanEdgeDetectionFilter(edgeFactor2); 
}

Frog: Filter Size 3×3

Frog: Filter Size 3x3

Sample Images

This article features a number of sample images. All featured images have been licensed allowing for reproduction. The following images feature as sample images:

Litoria_tyleri

Schrecklicherpfeilgiftfrosch-01

Dendropsophus_microcephalus_-_calling_male_(Cope,_1886)

Atelopus_zeteki1

Related Articles and Feedback

Feedback and questions are always encouraged. If you know of an alternative implementation or have ideas on a more efficient implementation please share in the comments section.

I’ve published a number of articles related to imaging and images of which you can find URL links here:

C# How to: Weighted Difference of Gaussians

Article Purpose

It is the purpose of this article to illustrate the concept of  . This article extends the conventional implementation of algorithms through the application of equally sized   only differing by a weight factor.

Frog: Kernel 5×5, Weight1 0.1, Weight2 2.1

Frog: Kernel 5x5, Weight1 0.1, Weight2 2.1

Sample Source Code

This article is accompanied by a sample source code Visual Studio project which is available for download here.

Using the Sample Application

This article relies on a sample application included as part of the accompanying sample source code. The sample application serves as a practical implementation of the concepts explored throughout this article.

The sample application user interface enables the user to configure and control the implementation of a filter. The configuration options exposed through the sample application’s user interface can be detailed as follows:

  • Load/Save Images – When executing the sample application users are able to load source/input from the local system through clicking the Load Image button. If desired, the sample application enables users to save resulting to the local file system through clicking the Save Image button.
  • Kernel Size – This option relates to the size of the that is to be implemented when performing through . Smaller are faster to compute and generally result in detected in the source/input to be expressed through thinner gradient edges. Larger can be computationally expensive to compute as sizes increase. In addition, the edges detected in source/input will generally be expressed as thicker gradient edges in resulting .
  • Weight Values – The sample application calculates   and in doing so implements a weight factor. A Weight Factor determines the blur intensity observed in result after having applied . Higher weight factors result in a more intense level of being applied. As expected, lower weight factors values result in  a less intense level of being applied. If the value of the first weight factor exceeds the value of the second weight factor resulting will be generated with a Black background and edges being indicated in White. In a similar fashion, when the second weight factor value exceeds that of the first weight factor resulting will be generated with a White background and edges being indicated in Black. The greater the difference between the first and second weight factor values result in a greater degree of removal. When weight factor values only differ slightly, resulting may be prone to .

The following image is screenshot of the Weighted Difference of Gaussians sample application in action:

Weighted Difference Of Gaussians Sample Application

Frog: Kernel 5×5, Weight1 1.8, Weight2 0.1

Frog: Kernel 5x5, Weight1 1.8, Weight2 0.1

Gaussian Blur

The algorithm can be described as one of the most popular and widely implemented methods of . From we gain the following excerpt:

A Gaussian blur (also known as Gaussian smoothing) is the result of blurring an image by a Gaussian function. It is a widely used effect in graphics software, typically to reduce image noise and reduce detail. The visual effect of this blurring technique is a smooth blur resembling that of viewing the image through a translucent screen, distinctly different from the bokeh effect produced by an out-of-focus lens or the shadow of an object under usual illumination. Gaussian smoothing is also used as a pre-processing stage in computer vision algorithms in order to enhance image structures at different scales.

Mathematically, applying a Gaussian blur to an image is the same as convolving the image with a Gaussian function. This is also known as a two-dimensional Weierstrass transform.

Take Note: The algorithm has the attribute of smoothing detail/definition whilst also having an edge preservation attribute. When applying a to an a level of detail/definition will be blurred/smoothed away, done in a fashion that would exclude/preserve edges.

Frog: Kernel 5×5, Weight1 2.7, Weight2 0.1

Frog: Kernel 5x5, Weight1 2.7, Weight2 0.1

Difference of Gaussians Edge Detection

refers to a specific method of . , common abbreviated as DoG, functions through the implementation of .

A clear and concise description can be found on the Wikipedia Article Page:

In imaging science, difference of Gaussians is a feature enhancement algorithm that involves the subtraction of one blurred version of an original image from another, less blurred version of the original. In the simple case of grayscale images, the blurred images are obtained by convolving the original grayscale images with Gaussian kernels having differing standard deviations. Blurring an image using a Gaussian kernel suppresses only high-frequency spatial information. Subtracting one image from the other preserves spatial information that lies between the range of frequencies that are preserved in the two blurred images. Thus, the difference of Gaussians is a band-pass filter that discards all but a handful of spatial frequencies that are present in the original grayscale image.

In a conventional sense involves applying to created as copies of the original source/input . There must be a difference in the size of the implemented when applying . A typical example would be applying a 3×3 on one copy whilst applying a 5×5 on another copy. The final step requires creating a result populated by subtracting the two blurred copies. The results obtained from subtraction represents the edges forming part of the source/input .

This article extends beyond the conventional method of implementing . The implementation illustrated in this article retains the core concept of subtracting values which have been blurred to different intensities. The implementation method explored here differs from the conventional method in the sense that the implemented do not differ in size. Both are in fact required to have the same size dimensions.

The implemented are equal in terms of their size dimensions, although values are different. Expressed from another angle: equally sized of which one represents a more intense level of than the other. A resulting intensity can be determined by the weight factor implemented when calculating the values.

Frog: Kernel 5×5, Weight1 3.7, Weight2 0.2

Frog: Kernel 5x5, Weight1 3.7, Weight2 0.2

The advantages of implementing equally sized can be described as follows:

Single Convolution implementation:  involves executing several nested code loops. Application performance can be severely negatively impacted when executing large nested loops. The conventional method of implementing generally involves having to implement two instances of , once per copy. The method implemented in this article executes the code loops related to only once. Considering the are equal in size, both can be iterated within the same set of loops.

Eliminating Image subtraction: In conventional implementations expressing differing intensity levels of have to be subtracted. The implementation method described in this article eliminates the need to perform subtraction. When applying using both simultaneously the two results obtained, one  from each , can be subtracted and assigned to the result . In addition, through calculating both results at the same time further reduces the need to create two temporary source copies.

Frog: Kernel 5×5, Weight1 2.4, Weight2 0.3

Frog: Kernel 5x5, Weight1 2.4, Weight2 0.3

Difference of Gaussians Edge Detection Required Steps

When implementing a several steps are required, those steps are detailed as follows:

  1. Calculate Kernels – Before implementing two have to be calculated. The calculated are required to be of equal size and differ in intensity. The sample application allows the user to configure intensity through updating the weight values, expressed as Weight 1 and Weight 2.
  2. Convert Source Image to Grayscale – Applying on   outperforms on RGB . When converting an RGB pixel to a pixel, colour components are combined to form a single gray level intensity. In other words a   consists of a third of the number of pixels when compared to the RGB from which the had been rendered. In the case of an ARGB the derived will be expressed in 25% of the number of pixels forming part of the source ARGB . When applying the number of processor cycles increases when the pixel count increases.
  3. Perform Convolution Implementing Thresholds – Using the newly created perform for both calculated in the first step. The result value equates to subtracting the two results obtained from .  If the result value exceeds the difference between the first and second weight value, the resulting pixel should be set to White, if not, set the result pixel to Black.

Frog: Kernel 5×5, Weight1 2.1, Weight2 0.5

Frog: Kernel 5x5, Weight1 2.1, Weight2 0.5

Calculating Gaussian Convolution Kernels

The sample application implements calculations. The implemented in are calculated at runtime, as opposed to being hard coded. Being able to dynamically construct has the advantage of providing a greater degree of control in runtime regarding application.

Several steps are involved in calculating . The first required step being to determine the Size and Weight. The size and weight factor of a comprises the two configurable values implemented when calculating . In the case of this article and the sample application those values will be configured by the user through the sample application’s user interface.

The formula implemented in calculating can be expressed as follows:

Gaussian Formula

The formula contains a number of symbols, which define how the filter will be implemented. The symbols forming part of the formula are described in the following list:

  • G(x y) – A value calculated using the Kernel formula. This value forms part of a , representing a single element.
  • π – Pi, one of the better known members of the Greek alphabet. The mathematical constant defined as 22 / 7.
  • σ – The lower case version of the Greek alphabet letter Sigma. This symbol simply represents a threshold or factor value, as specified by the user.
  • e – The formula references a lower case e symbol. The symbol represents . The value of has been defined as a mathematical constant equating to 2.71828182846.
  • x, y – The variables referenced as x and y relate to pixel coordinates within an . y Representing the vertical offset or row and x represents the horizontal offset or column.

Note: The formula’s implementation expects x and y to equal zero values when representing the coordinates of the pixel located in the middle of the .

Frog: Kernel 7×7, Weight1 0.1, Weight2 2.0

Frog: Kernel 7x7, Weight1 0.1, Weight2 2.0

Implementing Gaussian Kernel Calculations

The sample application defines the GaussianCalculator.Calculate method. This method accepts two parameters, kernel size and kernel weight. The following code snippet details the implementation:

public static double[,] Calculate(int lenght, double weight) 
{
    double[,] Kernel = new double [lenght, lenght]; 
    double sumTotal = 0; 

int kernelRadius = lenght / 2; double distance = 0;
double calculatedEuler = 1.0 / (2.0 * Math.PI * Math.Pow(weight, 2));
for (int filterY = -kernelRadius; filterY <= kernelRadius; filterY++) { for (int filterX = -kernelRadius; filterX <= kernelRadius; filterX++) { distance = ((filterX * filterX) + (filterY * filterY)) / (2 * (weight * weight));
Kernel[filterY + kernelRadius, filterX + kernelRadius] = calculatedEuler * Math.Exp(-distance);
sumTotal += Kernel[filterY + kernelRadius, filterX + kernelRadius]; } }
for (int y = 0; y < lenght; y++) { for (int x = 0; x < lenght; x++) { Kernel[y, x] = Kernel[y, x] * (1.0 / sumTotal); } }
return Kernel; }

Frog: Kernel 3×3, Weight1 0.1, Weight2 1.8

Frog: Kernel 3x3, Weight1 0.1, Weight2 1.8

Implementing Difference of Gaussians Edge Detection

The sample source code defines the DifferenceOfGaussianFilter method. This method has been defined as an targeting the class. The following code snippet provides the implementation:

public static Bitmap DifferenceOfGaussianFilter(this Bitmap sourceBitmap,  
                                                int matrixSize, double weight1, 
                                                double weight2) 
{
    double[,] kernel1 =  
    GaussianCalculator.Calculate(matrixSize,  
    (weight1 > weight2 ? weight1 : weight2)); 

double[,] kernel2 = GaussianCalculator.Calculate(matrixSize, (weight1 > weight2 ? weight2 : weight1));
BitmapData sourceData = sourceBitmap.LockBits(new Rectangle (0, 0, sourceBitmap.Width, sourceBitmap.Height), ImageLockMode.ReadOnly, PixelFormat.Format32bppArgb);
byte[] pixelBuffer = new byte [sourceData.Stride * sourceData.Height]; byte[] resultBuffer = new byte [sourceData.Stride * sourceData.Height]; byte[] grayscaleBuffer = new byte [sourceData.Width * sourceData.Height];
Marshal.Copy(sourceData.Scan0, pixelBuffer, 0, pixelBuffer.Length); sourceBitmap.UnlockBits(sourceData);
double rgb = 0;
for (int source = 0, dst = 0; source < pixelBuffer.Length && dst < grayscaleBuffer.Length; source += 4, dst++) { rgb = pixelBuffer * 0.11f; rgb += pixelBuffer * 0.59f; rgb += pixelBuffer * 0.3f;
grayscaleBuffer[dst] = (byte)rgb; }
double color1 = 0.0; double color2 = 0.0;
int filterOffset = (matrixSize - 1) / 2; int calcOffset = 0;
for (int source = 0, dst = 0; source < grayscaleBuffer.Length && dst + 4 < resultBuffer.Length; source++, dst += 4) { color1 = 0; color2 = 0;
for (int filterY = -filterOffset; filterY <= filterOffset; filterY++) { for (int filterX = -filterOffset; filterX <= filterOffset; filterX++) { calcOffset = source + (filterX) + (filterY * sourceBitmap.Width);
calcOffset = (calcOffset < 0 ? 0 : (calcOffset >= grayscaleBuffer.Length ? grayscaleBuffer.Length - 1 : calcOffset));
color1 += (grayscaleBuffer[calcOffset]) * kernel1[filterY + filterOffset, filterX + filterOffset];
color2 += (grayscaleBuffer[calcOffset]) * kernel2[filterY + filterOffset, filterX + filterOffset]; } }
color1 = color1 - color2; color1 = (color1 >= weight1 - weight2 ? 255 : 0);
resultBuffer[dst] = (byte)color1; resultBuffer[dst + 1] = (byte)color1; resultBuffer[dst + 2] = (byte)color1; resultBuffer[dst + 3] = 255; }
Bitmap resultBitmap = new Bitmap(sourceBitmap.Width, sourceBitmap.Height);
BitmapData resultData = resultBitmap.LockBits(new Rectangle (0, 0, resultBitmap.Width, resultBitmap.Height), ImageLockMode.WriteOnly, PixelFormat.Format32bppArgb);
Marshal.Copy(resultBuffer, 0, resultData.Scan0, resultBuffer.Length); resultBitmap.UnlockBits(resultData);
return resultBitmap; }

Frog: Kernel 3×3, Weight1 2.1, Weight2 0.7

Frog: Kernel 3x3, Weight1 2.1, Weight2 0.7

Sample Images

This article features a number of sample images. All featured images have been licensed allowing for reproduction. The following image files feature as sample images:

Panamanian Golden Frog

Panamanian Golden Frog

Dendropsophus Microcephalus

Dendropsophus Microcephalus

Tyler’s Tree Frog

Tyler's Tree Frog

Mimic Poison Frog

Mimic Poison Frog

Phyllobates Terribilis

Phyllobates Terribilis

Related Articles and Feedback

Feedback and questions are always encouraged. If you know of an alternative implementation or have ideas on a more efficient implementation please share in the comments section.

I’ve published a number of articles related to imaging and images of which you can find URL links here:

C# How to: Image Blur

Article Purpose

This article serves to provides an introduction and discussion relating to methods and techniques. The Image Blur methods covered in this article include: , , , and  .

Daisy: Mean 9×9

Daisy Mean 9x9

Sample Source Code

This article is accompanied by a sample source code Visual Studio project which is available for download .

Using the Sample Application

This article is accompanied by a sample application, intended to provide a means of testing and replicating topics discussed in this article. The sample application is a based application of which the user interface enables the user to select an type to implement.

When clicking the Load Image button users are able to browse the local file system in order to select source/input . In addition users are also able to save blurred result when clicking the Save Image button and browsing the local file system.

Daisy: Mean 7×7

Daisy Mean 7x7

The sample application provides the user with the ability to select the method of to implement. The dropdown located on the right-hand side of the user interface lists all of the supported methods of . When a user selects an item from the , the associated blur method will be implemented on the preview .

The image below is a screenshot of the Image Blur Filter sample application in action:

Image Blur Filter Sample Application

Image Blur Overview

The process of can be regarded as reducing the sharpness or crispness defined by an . results in detail/ being perceived as less distinct. are often blurred as a method of smoothing an .

perceived as too crisp/sharp can be softened by applying a variety of techniques and intensity levels. Often are smoothed/blurred in order to remove/reduce . In implementations better results are often achieved when first implementing through smoothing/. can even be implemented in a fashion where results reflect , a method known as .

In this article and the accompanying sample source code all methods of supported have been implemented through , with the exception of the filter. Each of the supported methods in essence only represent a different   . The technique capable of achieving optimal results will to varying degrees be dependent on the features present in the specified source/input . Each method provides a different set of desired properties and compromises. In the following sections an overview of each method will be discussed.

Daisy: Mean 9×9

Daisy Mean 9x9

Mean Filter/Box Blur

The also sometimes referred to as a represents a fairly simplistic implementation and definition. A definition can be found on as follows:

A box blur is an in which each pixel in the resulting image has a value equal to the average value of its neighbouring pixels in the input image. It is a form of low-pass ("blurring") filter and is a .

Due to its property of using equal weights it can be implemented using a much simpler accumulation algorithm which is significantly faster than using a sliding window algorithm.

as a title relates to all weight values in a being equal, therefore the alternate title of . In most cases a will only contain the value one. When performing implementing a , the factor value equates to the 1 being divided by the sum of all values.

The following is an example of a 5×5 convolution kernel:

Mean Filter Blur 5x5 Kernel

The consist of 25 elements, therefore the factor value equates to one divided by twenty five.

The Blur does not result in the same level of smoothing achieved by other methods. The method can also be susceptible to directional artefacts.

Daisy Mean 5×5

Daisy Mean 5x5

Gaussian Blur

The method of is a popular and often implemented filter. In contrast to the method produce resulting appearing to contain a more uniform level of smoothing. When implementing a is often applied to source/input resulting in . The has a good level of edge preservation, hence being used in operations.

From we gain the following :

A Gaussian blur (also known as Gaussian smoothing) is the result of blurring an image by a . It is a widely used effect in graphics software, typically to reduce image noise and reduce detail. The visual effect of this blurring technique is a smooth blur resembling that of viewing the image through a translucent screen, distinctly different from the bokeh effect produced by an out-of-focus lens or the shadow of an object under usual illumination. Gaussian smoothing is also used as a pre-processing stage in computer vision algorithms in order to enhance image structures at different scales

A potential drawback to implementing a results from the filter being computationally intensive. The following represents a 5×5 . The sum total of all elements in the equate to 159, therefore a factor value of 1.0 / 159.0 will be implemented.

Guassian Blur 5x5 Kernel

Daisy: Gaussian 5×5

Daisy Gaussian 5x5

Median Filter Blur

The is classified as a non-linear filter. In contrast to the other methods of discussed in this article the implementation does not involve or a predefined matrix . The following can be found on :

In signal processing, it is often desirable to be able to perform some kind of on an image or signal. The median filter is a nonlinear technique, often used to remove . Such noise reduction is a typical pre-processing step to improve the results of later processing (for example, on an image). Median filtering is very widely used in digital because, under certain conditions, it preserves edges while removing noise.

Daisy: Median 7×7

Daisy Median 7x7

As the name implies, the operates by calculating the value of a pixel group also referred to as a window. Calculating a value involves a number of steps. The required steps are listed as follows:

  1. Iterate each pixel that forms part of the source/input .
  2. In relation to the pixel currently being iterated determine neighbouring pixels located within the bounds defined by the window size. The window location should be offset in order to align the window’s middle pixel and the pixel currently being iterated.
  3. Neighbouring pixels located within the bounds  defined by the window should be added to a one dimensional neighbourhood array. Once all value have been added, the array should be sorted by value.
  4. The pixel value located at the middle of the sorted neighbourhood array qualifies as the value. The newly determined value should be assigned to the pixel currently being iterated.
  5. Repeat the steps listed above until all pixels within the source/input have been iterated.

Similar to the filter the has the ability to smooth whilst providing edge preservation. Depending on the window size implemented and the physical dimensions of input/source the can be computationally expensive.

Daisy: Median 9×9

Daisy Median 9x9

Motion Blur

The sample source implements filters. in the traditional sense has been association with photography and video capturing. can often be observed in scenarios where rapid movements are being captured to photographs or video recording. When recording a single frame, rapid movements could result in the changing  before the frame being captured has completed.

can be synthetically imitated through the implementation of Digital filters. The size of the provided when implementing affects the filter intensity perceived in result . Relating to filters the size of the specified in influences the perception and appearance of how rapidly movement had occurred to have blurred the resulting . Larger produce the appearance of more rapid motion, whereas smaller result in less rapid motion being perceived.

Daisy: Motion Blur 7×7 135 Degrees

Daisy Motion Blur 7x7 135 Degrees

Depending on the specified the ability exists to create the appearance of movement having occurred in a certain direction. The sample source code implements filters at 45 degrees, 135 degrees and in both directions simultaneously.

The listed below represents a 5×5 filter occurring at  45 degrees and 135 degrees:

MotionBlur5x5

Image Blur Implementation

The sample source code implements all of the concepts explored throughout this article. The source code definition can be grouped into 4 sections: ImageBlurFilter method, ConvolutionFilter method, MedianFilter method and the Matrix class. The following article sections relate to the 4 main source code sections.

The ImageBlurFilter has the purpose of invoking the correct blur filter method and relevant method parameters. This method acts as a method wrapper providing the technical implementation details required when performing a specified blur filter.

The definition of the ImageBlurFilter as follows:

 public static Bitmap ImageBlurFilter(this Bitmap sourceBitmap,  
                                             BlurType blurType) 
{  
     Bitmap resultBitmap = null; 

switch (blurType) { case BlurType.Mean3x3: { resultBitmap = sourceBitmap.ConvolutionFilter( Matrix.Mean3x3, 1.0 / 9.0, 0); } break; case BlurType.Mean5x5: { resultBitmap = sourceBitmap.ConvolutionFilter( Matrix.Mean5x5, 1.0 / 25.0, 0); } break; case BlurType.Mean7x7: { resultBitmap = sourceBitmap.ConvolutionFilter( Matrix.Mean7x7, 1.0 / 49.0, 0); } break; case BlurType.Mean9x9: { resultBitmap = sourceBitmap.ConvolutionFilter( Matrix.Mean9x9, 1.0 / 81.0, 0); } break; case BlurType.GaussianBlur3x3: { resultBitmap = sourceBitmap.ConvolutionFilter( Matrix.GaussianBlur3x3, 1.0 / 16.0, 0); } break; case BlurType.GaussianBlur5x5: { resultBitmap = sourceBitmap.ConvolutionFilter( Matrix.GaussianBlur5x5, 1.0 / 159.0, 0); } break; case BlurType.MotionBlur5x5: { resultBitmap = sourceBitmap.ConvolutionFilter( Matrix.MotionBlur5x5, 1.0 / 10.0, 0); } break; case BlurType.MotionBlur5x5At45Degrees: { resultBitmap = sourceBitmap.ConvolutionFilter( Matrix.MotionBlur5x5At45Degrees, 1.0 / 5.0, 0); } break; case BlurType.MotionBlur5x5At135Degrees: { resultBitmap = sourceBitmap.ConvolutionFilter( Matrix.MotionBlur5x5At135Degrees, 1.0 / 5.0, 0); } break; case BlurType.MotionBlur7x7: { resultBitmap = sourceBitmap.ConvolutionFilter( Matrix.MotionBlur7x7, 1.0 / 14.0, 0); } break; case BlurType.MotionBlur7x7At45Degrees: { resultBitmap = sourceBitmap.ConvolutionFilter( Matrix.MotionBlur7x7At45Degrees, 1.0 / 7.0, 0); } break; case BlurType.MotionBlur7x7At135Degrees: { resultBitmap = sourceBitmap.ConvolutionFilter( Matrix.MotionBlur7x7At135Degrees, 1.0 / 7.0, 0); } break; case BlurType.MotionBlur9x9: { resultBitmap = sourceBitmap.ConvolutionFilter( Matrix.MotionBlur9x9, 1.0 / 18.0, 0); } break; case BlurType.MotionBlur9x9At45Degrees: { resultBitmap = sourceBitmap.ConvolutionFilter( Matrix.MotionBlur9x9At45Degrees, 1.0 / 9.0, 0); } break; case BlurType.MotionBlur9x9At135Degrees: { resultBitmap = sourceBitmap.ConvolutionFilter( Matrix.MotionBlur9x9At135Degrees, 1.0 / 9.0, 0); } break; case BlurType.Median3x3: { resultBitmap = sourceBitmap.MedianFilter(3); } break; case BlurType.Median5x5: { resultBitmap = sourceBitmap.MedianFilter(5); } break; case BlurType.Median7x7: { resultBitmap = sourceBitmap.MedianFilter(7); } break; case BlurType.Median9x9: { resultBitmap = sourceBitmap.MedianFilter(9); } break; case BlurType.Median11x11: { resultBitmap = sourceBitmap.MedianFilter(11); } break; }
return resultBitmap; }

Daisy: Motion Blur 9×9

Daisy Motion Blur 9x9

The Matrix class serves as a collection of  various definitions. The Matrix class and all public properties are defined as static. The definition of the Matrix class as follows:

     public static class Matrix 
    {  
         public static double[,] Mean3x3 
         {  
             get 
             {  
                 return new double[,]   
                { {  1, 1, 1, },  
                  {  1, 1, 1, },  
                  {  1, 1, 1, }, }; 
             }  
         }  

public static double[,] Mean5x5 { get { return new double[,] { { 1, 1, 1, 1, 1 }, { 1, 1, 1, 1, 1 }, { 1, 1, 1, 1, 1 }, { 1, 1, 1, 1, 1 }, { 1, 1, 1, 1, 1 }, }; } }
public static double[,] Mean7x7 { get { return new double[,] { { 1, 1, 1, 1, 1, 1, 1 }, { 1, 1, 1, 1, 1, 1, 1 }, { 1, 1, 1, 1, 1, 1, 1 }, { 1, 1, 1, 1, 1, 1, 1 }, { 1, 1, 1, 1, 1, 1, 1 }, { 1, 1, 1, 1, 1, 1, 1 }, { 1, 1, 1, 1, 1, 1, 1 }, }; } }
public static double[,] Mean9x9 { get { return new double[,] { { 1, 1, 1, 1, 1, 1, 1, 1, 1 }, { 1, 1, 1, 1, 1, 1, 1, 1, 1 }, { 1, 1, 1, 1, 1, 1, 1, 1, 1 }, { 1, 1, 1, 1, 1, 1, 1, 1, 1 }, { 1, 1, 1, 1, 1, 1, 1, 1, 1 }, { 1, 1, 1, 1, 1, 1, 1, 1, 1 }, { 1, 1, 1, 1, 1, 1, 1, 1, 1 }, { 1, 1, 1, 1, 1, 1, 1, 1, 1 }, { 1, 1, 1, 1, 1, 1, 1, 1, 1 }, }; } }
public static double[,] GaussianBlur3x3 { get { return new double[,] { { 1, 2, 1, }, { 2, 4, 2, }, { 1, 2, 1, }, }; } }
public static double[,] GaussianBlur5x5 { get { return new double[,] { { 2, 04, 05, 04, 2 }, { 4, 09, 12, 09, 4 }, { 5, 12, 15, 12, 5 }, { 4, 09, 12, 09, 4 }, { 2, 04, 05, 04, 2 }, }; } }
public static double[,] MotionBlur5x5 { get { return new double[,] { { 1, 0, 0, 0, 1 }, { 0, 1, 0, 1, 0 }, { 0, 0, 1, 0, 0 }, { 0, 1, 0, 1, 0 }, { 1, 0, 0, 0, 1 }, }; } }
public static double[,] MotionBlur5x5At45Degrees { get { return new double[,] { { 0, 0, 0, 0, 1 }, { 0, 0, 0, 1, 0 }, { 0, 0, 1, 0, 0 }, { 0, 1, 0, 0, 0 }, { 1, 0, 0, 0, 0 }, }; } }
public static double[,] MotionBlur5x5At135Degrees { get { return new double[,] { { 1, 0, 0, 0, 0 }, { 0, 1, 0, 0, 0 }, { 0, 0, 1, 0, 0 }, { 0, 0, 0, 1, 0 }, { 0, 0, 0, 0, 1 }, }; } }
public static double[,] MotionBlur7x7 { get { return new double[,] { { 1, 0, 0, 0, 0, 0, 1 }, { 0, 1, 0, 0, 0, 1, 0 }, { 0, 0, 1, 0, 1, 0, 0 }, { 0, 0, 0, 1, 0, 0, 0 }, { 0, 0, 1, 0, 1, 0, 0 }, { 0, 1, 0, 0, 0, 1, 0 }, { 1, 0, 0, 0, 0, 0, 1 }, }; } }
public static double[,] MotionBlur7x7At45Degrees { get { return new double[,] { { 0, 0, 0, 0, 0, 0, 1 }, { 0, 0, 0, 0, 0, 1, 0 }, { 0, 0, 0, 0, 1, 0, 0 }, { 0, 0, 0, 1, 0, 0, 0 }, { 0, 0, 1, 0, 0, 0, 0 }, { 0, 1, 0, 0, 0, 0, 0 }, { 1, 0, 0, 0, 0, 0, 0 }, }; } }
public static double[,] MotionBlur7x7At135Degrees { get { return new double[,] { { 1, 0, 0, 0, 0, 0, 0 }, { 0, 1, 0, 0, 0, 0, 0 }, { 0, 0, 1, 0, 0, 0, 0 }, { 0, 0, 0, 1, 0, 0, 0 }, { 0, 0, 0, 0, 1, 0, 0 }, { 0, 0, 0, 0, 0, 1, 0 }, { 0, 0, 0, 0, 0, 0, 1 }, }; } }
public static double[,] MotionBlur9x9 { get { return new double[,] { { 1, 0, 0, 0, 0, 0, 0, 0, 1, }, { 0, 1, 0, 0, 0, 0, 0, 1, 0, }, { 0, 0, 1, 0, 0, 0, 1, 0, 0, }, { 0, 0, 0, 1, 0, 1, 0, 0, 0, }, { 0, 0, 0, 0, 1, 0, 0, 0, 0, }, { 0, 0, 0, 1, 0, 1, 0, 0, 0, }, { 0, 0, 1, 0, 0, 0, 1, 0, 0, }, { 0, 1, 0, 0, 0, 0, 0, 1, 0, }, { 1, 0, 0, 0, 0, 0, 0, 0, 1, }, }; } }
public static double[,] MotionBlur9x9At45Degrees { get { return new double[,] { { 0, 0, 0, 0, 0, 0, 0, 0, 1, }, { 0, 0, 0, 0, 0, 0, 0, 1, 0, }, { 0, 0, 0, 0, 0, 0, 1, 0, 0, }, { 0, 0, 0, 0, 0, 1, 0, 0, 0, }, { 0, 0, 0, 0, 1, 0, 0, 0, 0, }, { 0, 0, 0, 1, 0, 0, 0, 0, 0, }, { 0, 0, 1, 0, 0, 0, 0, 0, 0, }, { 0, 1, 0, 0, 0, 0, 0, 0, 0, }, { 1, 0, 0, 0, 0, 0, 0, 0, 0, }, }; } }
public static double[,] MotionBlur9x9At135Degrees { get { return new double[,] { { 1, 0, 0, 0, 0, 0, 0, 0, 0, }, { 0, 1, 0, 0, 0, 0, 0, 0, 0, }, { 0, 0, 1, 0, 0, 0, 0, 0, 0, }, { 0, 0, 0, 1, 0, 0, 0, 0, 0, }, { 0, 0, 0, 0, 1, 0, 0, 0, 0, }, { 0, 0, 0, 0, 0, 1, 0, 0, 0, }, { 0, 0, 0, 0, 0, 0, 1, 0, 0, }, { 0, 0, 0, 0, 0, 0, 0, 1, 0, }, { 0, 0, 0, 0, 0, 0, 0, 0, 1, }, }; } } }

Daisy: Median 7×7

Daisy Median 7x7

The MedianFilter targets the class. The MedianFilter method applies a using the specified and matrix size (window size), returning a new representing the filtered .

The definition of the MedianFilter as follows:

 public static Bitmap MedianFilter(this Bitmap sourceBitmap, 
                                   int matrixSize) 
{ 
     BitmapData sourceData = 
                sourceBitmap.LockBits(new Rectangle(0, 0, 
                sourceBitmap.Width, sourceBitmap.Height), 
                ImageLockMode.ReadOnly, 
                PixelFormat.Format32bppArgb); 

byte[] pixelBuffer = new byte[sourceData.Stride * sourceData.Height];
byte[] resultBuffer = new byte[sourceData.Stride * sourceData.Height];
Marshal.Copy(sourceData.Scan0, pixelBuffer, 0, pixelBuffer.Length);
sourceBitmap.UnlockBits(sourceData);
int filterOffset = (matrixSize - 1) / 2; int calcOffset = 0;
int byteOffset = 0;
List<int> neighbourPixels = new List<int>(); byte[] middlePixel;
for (int offsetY = filterOffset; offsetY < sourceBitmap.Height - filterOffset; offsetY++) { for (int offsetX = filterOffset; offsetX < sourceBitmap.Width - filterOffset; offsetX++) { byteOffset = offsetY * sourceData.Stride + offsetX * 4;
neighbourPixels.Clear();
for (int filterY = -filterOffset; filterY <= filterOffset; filterY++) { for (int filterX = -filterOffset; filterX <= filterOffset; filterX++) {
calcOffset = byteOffset + (filterX * 4) + (filterY * sourceData.Stride);
neighbourPixels.Add(BitConverter.ToInt32( pixelBuffer, calcOffset)); } }
neighbourPixels.Sort(); middlePixel = BitConverter.GetBytes( neighbourPixels[filterOffset]);
resultBuffer[byteOffset] = middlePixel[0]; resultBuffer[byteOffset + 1] = middlePixel[1]; resultBuffer[byteOffset + 2] = middlePixel[2]; resultBuffer[byteOffset + 3] = middlePixel[3]; } }
Bitmap resultBitmap = new Bitmap (sourceBitmap.Width, sourceBitmap.Height);
BitmapData resultData = resultBitmap.LockBits(new Rectangle (0, 0, resultBitmap.Width, resultBitmap.Height), ImageLockMode.WriteOnly, PixelFormat.Format32bppArgb);
Marshal.Copy(resultBuffer, 0, resultData.Scan0, resultBuffer.Length);
resultBitmap.UnlockBits(resultData);
return resultBitmap; }

Daisy: Motion Blur 9×9

Daisy Motion Blur 9x9

The sample source code performs by invoking the ConvolutionFilter .

The definition of the ConvolutionFilter as follows:

private static Bitmap ConvolutionFilter(this Bitmap sourceBitmap, 
                                          double[,] filterMatrix, 
                                               double factor = 1, 
                                                    int bias = 0) 
{ 
    BitmapData sourceData = sourceBitmap.LockBits(new Rectangle(0, 0, 
                             sourceBitmap.Width, sourceBitmap.Height), 
                                               ImageLockMode.ReadOnly, 
                                         PixelFormat.Format32bppArgb); 

byte[] pixelBuffer = new byte[sourceData.Stride * sourceData.Height]; byte[] resultBuffer = new byte[sourceData.Stride * sourceData.Height];
Marshal.Copy(sourceData.Scan0, pixelBuffer, 0, pixelBuffer.Length); sourceBitmap.UnlockBits(sourceData);
double blue = 0.0; double green = 0.0; double red = 0.0;
int filterWidth = filterMatrix.GetLength(1); int filterHeight = filterMatrix.GetLength(0);
int filterOffset = (filterWidth - 1) / 2; int calcOffset = 0;
int byteOffset = 0;
for (int offsetY = filterOffset; offsetY < sourceBitmap.Height - filterOffset; offsetY++) { for (int offsetX = filterOffset; offsetX < sourceBitmap.Width - filterOffset; offsetX++) { blue = 0; green = 0; red = 0;
byteOffset = offsetY * sourceData.Stride + offsetX * 4;
for (int filterY = -filterOffset; filterY <= filterOffset; filterY++) { for (int filterX = -filterOffset; filterX <= filterOffset; filterX++) {
calcOffset = byteOffset + (filterX * 4) + (filterY * sourceData.Stride);
blue += (double)(pixelBuffer[calcOffset]) * filterMatrix[filterY + filterOffset, filterX + filterOffset];
green += (double)(pixelBuffer[calcOffset + 1]) * filterMatrix[filterY + filterOffset, filterX + filterOffset];
red += (double)(pixelBuffer[calcOffset + 2]) * filterMatrix[filterY + filterOffset, filterX + filterOffset]; } }
blue = factor * blue + bias; green = factor * green + bias; red = factor * red + bias;
blue = (blue > 255 ? 255 : (blue < 0 ? 0 : blue));
green = (green > 255 ? 255 : (green < 0 ? 0 : green));
red = (red > 255 ? 255 : (red < 0 ? 0 : red));
resultBuffer[byteOffset] = (byte)(blue); resultBuffer[byteOffset + 1] = (byte)(green); resultBuffer[byteOffset + 2] = (byte)(red); resultBuffer[byteOffset + 3] = 255; } }
Bitmap resultBitmap = new Bitmap(sourceBitmap.Width, sourceBitmap.Height);
BitmapData resultData = resultBitmap.LockBits(new Rectangle (0, 0, resultBitmap.Width, resultBitmap.Height), ImageLockMode.WriteOnly, PixelFormat.Format32bppArgb);
Marshal.Copy(resultBuffer, 0, resultData.Scan0, resultBuffer.Length); resultBitmap.UnlockBits(resultData);
return resultBitmap; }

Sample Images

This article features a number of sample images. All featured images have been licensed allowing for reproduction.

The sample images featuring an image of a yellow daisy is licensed under the Creative Commons Attribution-Share Alike 2.5 Generic license and can be downloaded from Wikimedia.org.

The sample images featuring an image of a white daisy is licensed under the Creative Commons Attribution-Share Alike 3.0 Unported license and can be downloaded from Wikipedia.

The sample images featuring an image of a pink daisy is licensed under the Creative Commons Attribution-Share Alike 2.5 Generic license and can be downloaded from Wikipedia.

The sample images featuring an image of a purple daisy is licensed under the Creative Commons Attribution-ShareAlike 3.0 License and can be downloaded from Wikipedia.

The Original Image

Purple_osteospermum

Daisy: Gaussian 3×3

Daisy Gaussian 3x3

Daisy: Gaussian 5×5

Daisy Gaussian 5x5

Daisy: Mean 3×3

Daisy Mean 3x3

Daisy: Mean 5×5

Daisy Mean 5x5

Daisy: Mean 7×7

Daisy Mean 7x7

Daisy: Mean 9×9

Daisy Mean 9x9

Daisy: Median 3×3

Daisy Median 3x3

Daisy: Median 5×5

Daisy Median 5x5

Daisy: Median 7×7

Daisy Median 7x7

Daisy: Median 9×9

Daisy Median 9x9

Daisy: Median 11×11

Daisy Median 11x11

Daisy: Motion Blur 5×5

Daisy Motion Blur 5x5

Daisy: Motion Blur 5×5 45 Degrees

Daisy Motion Blur 5x5 45 Degrees

Daisy: Motion Blur 5×5 135 Degrees

Daisy Motion Blur 5x5 135 Degrees

Daisy: Motion Blur 7×7

Daisy Motion Blur 7x7

Daisy: Motion Blur 7×7 45 Degrees

Daisy Motion Blur 7x7 45 Degree

Daisy: Motion Blur 7×7 135 Degrees

Daisy Motion Blur 7x7 135 Degrees

Daisy: Motion Blur 9×9

Daisy Motion Blur 9x9

Daisy: Motion Blur 9×9 45 Degrees

Daisy Motion Blur 9x9 45 Degrees

Daisy: Motion Blur 9×9 135 Degrees

Daisy Motion Blur 9x9 135 Degrees

Related Articles and Feedback

Feedback and questions are always encouraged. If you know of an alternative implementation or have ideas on a more efficient implementation please share in the comments section.

I’ve published a number of articles related to imaging and images of which you can find URL links here:

C# How to: Calculating Gaussian Kernels

Article Purpose

This purpose of this article is to explain and illustrate in detail the requirements involved in calculating intended for use in when implementing filters. This article’s discussion spans from exploring concepts in theory and continues on to implement concepts through C# sample source code.

Ant: Gaussian Kernel 5×5 Weight 19

Ant Gaussian Kernel 5x5 Weight 19

Sample Source Code

This article is accompanied by a sample source code Visual Studio project which is available for download

Using the Sample Application

A Sample Application forms part of the accompanying sample source code, intended to implement the topics discussed and also provides the means to replicate and test the concepts being illustrated.

The sample application is a based application which provides functionality enabling users to generate/calculate  . Calculation results are influenced through user specified options in the form of: Kernel Size and Weight.

Ladybird: Gaussian Kernel 5×5 Weight 5.5

Gaussian Kernel 5x5 Weight 5.5

In the sample application and related sample source code when referring to Kernel Size, a reference is being made relating to the physical size dimensions of the / used in . When higher values are specified in setting the Kernel Size, the resulting output will reflect a greater degree of . Kernel Sizes being specified as lower values result in the output reflecting a lesser degree of .

In a similar fashion to the Kernel size value, the Weight value provided when generating a results in smoother/more blurred when specified as higher values. Lower values assigned to the Weight value has the expected result of less being evident in output .

Prey Mantis: Gaussian Kernel 13×13 Weight 13

Prey Mantis Gaussian Kernel 13x13 Weight 13

The sample application has the ability to provide the user with a visual representation implementing the calculated value . Users are able to select source/input from the local system by clicking the Load Image button. When desired, users are able to save blurred/filtered to the local file system by clicking the Save Image button.

The image below is screenshot of the Gaussian Kernel Calculator sample application in action:

Gaussian Kernel Calculator Sample Application

Calculating Gaussian Convolution Kernels

The formula implemented in calculating can be implemented in C# source code fairly easily. Once the method in which the formula operates has been grasped the actual code implementation becomes straight forward.

The formula can be expressed as follows:

Gaussian Kernel formula

The formula contains a number of symbols, which define how the filter will be implemented. The symbols forming part of the formula are described in the following list:

  • G(x y) – A value calculated using the formula. This value forms part of a , representing a single element.
  • π – Pi, one of the better known members of the Greek alphabet. The mathematical constant defined as 22 / 7.
  • σ – The lower case version of the Greek alphabet letter Sigma. This symbol simply represents a threshold or factor value, as specified by the user.
  • e – The formula references a lower case e symbol. The symbol represents . The value of has been defined as a mathematical constant equating to 2.71828182846.
  • x, y – The variables referenced as x and y relate to pixel coordinates within an . y Representing the vertical offset or row and x represents the horizontal offset or column.

Note: The formula’s implementation expects x and y to equal zero values when representing the coordinates of the pixel located in the middle of the .

Ladybird: Gaussian Kernel 13×13 Weight 9.5

Gaussian Kernel 13x13 Weight 9.5

When calculating the elements, the coordinate values expressed by x and y should reflect the distance in pixels from the middle pixel. All coordinate values must be greater than zero.

In order to gain a better grasp on the formula we can implement the formula in steps. If we were to create a 3×3 and specified a weighting value of 5.5 our calculations can start off as indicated by the following illustration:

Gaussian Kernel Formula

The formula has been implement on each element forming part of the , 9 values in total. Coordinate values have now been replaced with actual values, differing for each position/element. Calculating zero to the power of two equates to zero. In the scenario above indicating zeros which express exponential values might help to ease initial understanding, as opposed to providing simplified values and potentially causing confusing scenarios. The following image illustrates the calculated values of each element:

Gaussian Kernel Values non summed

Ant: Gaussian Kernel 9×9 Weight 19

Ant Gaussian Kernel 9x9 Weight 19

An important requirement to take note of at this point being that the sum total of all the elements contained as part of a / must equate to one. Looking at our calculated results that is not the case. The needs to be modified in order to satisfy the requirement of having a sum total value of 1 when adding together all the elements of the .

At this point the sum total of the equates to 0.046322548968. We can correct the values, ensuring the sum total of all elements equate to 1. The values should be updated by multiplying each element by one divided by the current sum. In other words each item should be multiplied by:

1.0 / 0.046322548968

After updating the by multiplying each element with the values mentioned above, the result as follows:

Calculated Gaussian Kernel Values

We have now successfully calculated a 3×3 which implements a weight value of 5.5. Implementing the has the following effect:

Rose: Gaussian Kernel 3×3 Weight 5.5

Rose Gaussian Kernel 3x3 Weight 5.5

The Original Image

Rose_Amber_Flush_20070601

The calculated can now be implemented when performing .

Implementing Gaussian Kernel Calculations

In this section of the article we will be exploring how to implement kernel calculations in terms of C# code. Defined as part of the sample source code the of the static MatrixCalculator class, exposing the static Calculate method. All of the formula calculation tasks discussed in the previous section have been implemented within this method.

As parameter values the method expects a value indicating the size and a value representing the Weight value. The Calculate method returns a two dimensional array of type double. The return value array represents the calculated .

The definition of the MatrixCalculator.Calculate method as follows:

public static double[,] Calculate(int length, double weight) 
{
    double[,] Kernel = new double [length, lenght]; 
    double sumTotal = 0; 

  
    int kernelRadius = lenght / 2; 
    double distance = 0; 

  
    double calculatedEuler = 1.0 /  
    (2.0 * Math.PI * Math.Pow(weight, 2)); 

  
    for (int filterY = -kernelRadius; 
         filterY <= kernelRadius; filterY++) 
    {
        for (int filterX = -kernelRadius; 
            filterX <= kernelRadius; filterX++) 
        {
            distance = ((filterX * filterX) +  
                       (filterY * filterY)) /  
                       (2 * (weight * weight)); 

  
            Kernel[filterY + kernelRadius,  
                   filterX + kernelRadius] =  
                   calculatedEuler * Math.Exp(-distance); 

  
            sumTotal += Kernel[filterY + kernelRadius,  
                               filterX + kernelRadius]; 
        } 
    } 

  
    for (int y = 0; y < lenght; y++) 
    { 
        for (int x = 0; x < lenght; x++) 
        { 
            Kernel[y, x] = Kernel[y, x] *  
                           (1.0 / sumTotal); 
        } 
    } 

  
    return Kernel; 
}

Ladybird: Gaussian Kernel 19×19 Weight 9.5

Gaussian Kernel 19x19 Weight 9.5

The sample source code provides the definition of the ConvolutionFilter , targeting the class. This method accepts as a parameter a two dimensional array representing the to implement when performing . The value passed to this function originates from the calculated .

Detailed below is the definition of the ConvolutionFilter :

public static Bitmap ConvolutionFilter(this Bitmap sourceBitmap,  
                                         double[,] filterMatrix,  
                                              double factor = 1,  
                                                   int bias = 0)  
{ 
    BitmapData sourceData = sourceBitmap.LockBits(new Rectangle(0, 0, 
                            sourceBitmap.Width, sourceBitmap.Height), 
                                              ImageLockMode.ReadOnly,  
                                        PixelFormat.Format32bppArgb); 

   
    byte[] pixelBuffer = new byte[sourceData.Stride * sourceData.Height]; 
    byte[] resultBuffer = new byte[sourceData.Stride * sourceData.Height]; 

   
    Marshal.Copy(sourceData.Scan0, pixelBuffer, 0, pixelBuffer.Length); 
    sourceBitmap.UnlockBits(sourceData); 

   
    double blue = 0.0; 
    double green = 0.0; 
    double red = 0.0; 

   
    int filterWidth = filterMatrix.GetLength(1); 
    int filterHeight = filterMatrix.GetLength(0); 

   
    int filterOffset = (filterWidth-1) / 2; 
    int calcOffset = 0; 

   
    int byteOffset = 0; 

   
    for (int offsetY = filterOffset; offsetY <  
        sourceBitmap.Height - filterOffset; offsetY++) 
    {
        for (int offsetX = filterOffset; offsetX <  
            sourceBitmap.Width - filterOffset; offsetX++) 
        {
            blue = 0; 
            green = 0; 
            red = 0; 

   
            byteOffset = offsetY *  
                         sourceData.Stride +  
                         offsetX * 4; 

   
            for (int filterY = -filterOffset;  
                filterY <= filterOffset; filterY++) 
            { 
                for (int filterX = -filterOffset; 
                    filterX <= filterOffset; filterX++) 
                { 

   
                    calcOffset = byteOffset +  
                                 (filterX * 4) +  
                                 (filterY * sourceData.Stride); 

   
                    blue += (double  )(pixelBuffer[calcOffset]) * 
                            filterMatrix[filterY + filterOffset,  
                                                filterX + filterOffset]; 

   
                    green += (double  )(pixelBuffer[calcOffset + 1]) * 
                             filterMatrix[filterY + filterOffset,  
                                                filterX + filterOffset]; 

   
                    red += (double  )(pixelBuffer[calcOffset + 2]) * 
                           filterMatrix[filterY + filterOffset,  
                                              filterX + filterOffset]; 
                } 
            } 

   
            blue = factor * blue + bias; 
            green = factor * green + bias; 
            red = factor * red + bias; 

   
            blue = (blue > 255 ? 255 : (blue < 0 ? 0 : blue)); 
            green = (green > 255 ? 255 : (green < 0 ? 0 : green)); 
            red = (red > 255 ? 255 : (red < 0 ? 0 : blue)); 

   
            resultBuffer[byteOffset] = (byte)(blue); 
            resultBuffer[byteOffset + 1] = (byte)(green); 
            resultBuffer[byteOffset + 2] = (byte)(red); 
            resultBuffer[byteOffset + 3] = 255; 
        }
    }

   
    Bitmap resultBitmap = new Bitmap(sourceBitmap.Width, sourceBitmap.Height); 

   
    BitmapData resultData = resultBitmap.LockBits(new Rectangle (0, 0, 
                             resultBitmap.Width, resultBitmap.Height), 
                                              ImageLockMode.WriteOnly, 
                                         PixelFormat.Format32bppArgb); 

   
    Marshal.Copy(resultBuffer, 0, resultData.Scan0, resultBuffer.Length); 
    resultBitmap.UnlockBits(resultData); 

   
    return resultBitmap; 
}

Ant: Gaussian Kernel 7×7 Weight 19

Ant Gaussian Kernel 7x7 Weight 19

Sample Images

This article features a number of sample images. All featured images have been licensed allowing for reproduction.

The sample images featuring an image of a prey mantis is licensed under the Creative Commons Attribution-Share Alike 3.0 Unported license and can be downloaded from .

The sample images featuring an image of an ant has been released into the public domain by its author, Sean.hoyland. This applies worldwide. In some countries this may not be legally possible; if so: Sean.hoyland grants anyone the right to use this work for any purpose, without any conditions, unless such conditions are required by law. The original image can be downloaded from .

The sample images featuring an image of a ladybird (ladybug or lady beetle) is licensed under the Creative Commons Attribution-Share Alike 2.0 Generic license and can be downloaded from Wikipedia.

The sample images featuring an image of a wasp is licensed under the Creative Commons Attribution-Share Alike 3.0 Unported license and can be downloaded from Wikimedia.org.

The Original Image

1280px-Gemeiner_Widderbock_4483

Wasp Gaussian Kernel 3×3 Weight 9.25

Wasp Gaussian Kernel 3x3 Weight 9.25

Wasp Gaussian Kernel 5×5 Weight 9.25

Wasp Gaussian Kernel 5x5 Weight 9.25

Wasp Gaussian Kernel 7×7 Weight 9.25

Wasp Gaussian Kernel 7x7 Weight 9.25

Wasp Gaussian Kernel 9×9 Weight 9.25

Wasp Gaussian Kernel 9x9 Weight 9.25

Wasp Gaussian Kernel 11×11 Weight 9.25

Wasp Gaussian Kernel 11x11 Weight 9.25

Wasp Gaussian Kernel 13×13 Weight 9.25

Wasp Gaussian Kernel 13x13 Weight 9.25

Wasp Gaussian Kernel 15×15 Weight 9.25

Wasp Gaussian Kernel 15x15 Weight 9.25

Wasp Gaussian Kernel 17×17 Weight 9.25

Wasp Gaussian Kernel 17x17 Weight 9.25

Wasp Gaussian Kernel 19×19 Weight 9.25

Wasp Gaussian Kernel 19x19 Weight 9.25

Related Articles and Feedback

Feedback and questions are always encouraged. If you know of an alternative implementation or have ideas on a more efficient implementation please share in the comments section.

I’ve published a number of articles related to imaging and images of which you can find URL links here:


Dewald Esterhuizen

Blog Stats

  • 435,579 hits

Enter your email address to follow and receive notifications of new posts by email.

Join 204 other followers

Archives

Twitter feed


%d bloggers like this: