Photoshop algorithm implementation Advanced-blur filter-Gaussian filter (23)

Photoshop Algorithm Implementation Advanced-blur filter-Gaussian filter (23)

[Email protected]

Http://blog.csdn.net/kezunhai

Gaussian Blur (Gaus Blur) uses a two-Gaussian template to blur the image, which is used for image blurring (removal of detail and noise), and its processing effect gives a better feeling of softness.

One Gaussian:

Ivigos:

Theoretically, Gaussian distributions have non-negative values on all domain definitions, which requires an infinitely large convolution core. In fact, you only need to take the value of 3 times times the standard deviation around the mean value, and remove it directly from the external part. As a standard deviation of 1.0 integer value of the Gaussian core:

In some cases, the following modules can be used to approximate the Gaussian core:

Based on the kernel size and variance of the user input, the Gaussian kernel is computed and Gaussian filter is realized, and the Gaussian kernel creation algorithm is as follows:

Create a Gaussian core///ksize: The size of the convolution core 3, 5, 7, etc. (3x3, 5x5, 7x7)/sigma: Variance void Photoshop::creategausskernel (float **pdkernel, int ksize, float sigma) {float DDIs;  A point in the array to the center point distance float dcenter = (kSize-1)/2;float dSum = 0;  Float Dvalue;int I, J; Generate Gaussian data for  (i=0; i<ksize; i++) {for (j=0; j<ksize; J + +) {DDIs = FABSF (i-dcenter) + FABSF (j-dcenter);  Dvalue = exp (-ddis*ddis/(2*sigma*sigma+dbl_epsilon));pd kernel[i][j] = Dvalue with sum to approximate square sum;  Dsum+=dvalue; }}//Normalization for  (i=0; i<ksize; i++) {for (j=0; j<ksize; J + +) {Pdkernel[i][j]/= (Dsum+dbl_epsilon);}}  

Gaussian filter algorithm:

void PhotoShop :: Gaussianblur (Cv::mat &img, Cv::mat &dst, int ksize, float sigma) {int height = img.rows;int width = img.cols;int Chns = Img.channels (); int I, J, K,p, q;//allocate Gaussian kernel space float **pkernel = new Float*[ksize];for (i=0; i<ksize; i++) {pkernel[i ] = new Float[ksize]; ZeroMemory (Pkernel[i], ksize*sizeof (float));}  Creategausskernel (Pkernel, Ksize, Sigma); Create a Gaussian kernel if (Dst.empty ()) dst.create (height, width, img.type ()), float sum = 0;int border = (kSize-1)/2;for (i =border; i&l T;height-border; i++) {unsigned char* dstdata = (unsigned char*) Dst.data + dst.step*i;for (j=border; j<width-border; J + +) {for (k=0; K&L T;chns; k++) {sum = 0;for (p =-border; p<=border; p++) {for (q =-border; q<=border; q++) {sum + = GetPixel (img, i+p, j+q, K) *pkernel[p+border][q+border];}} Dstdata[j*chns+k] = saturate_cast<uchar> (sum+0.5);} For-k}}//For i} 

Algorithm test:

Gaussian filter has been widely used in many algorithms because of its linearity and good denoising effect, such as the Gaussian filter used in preprocessing of canny operator, and the implementation of Sobel operator in OpenCV is also used as preprocessing step, It's still famous. The SIFT operator uses Gaussian pyramid and Gaussian difference to extract candidate feature points, and interested friends can further understand the application of Gaussian filtering.

Kezunhai Source:Http://blog.csdn.net/kezunhai Welcome to reprint or share, but be sure to declare the source of the article.

Photoshop algorithm implementation Advanced-blur filter-Gaussian filter (23)