High-efficiency median filtering (using Copymakeborder to process boundary pixels)

The median filtering algorithm in the previous filter is modified, and the boundary pixels are processed by copymakeborder.

void Medianblur (const mat &AMP;IMAGE_SRC, Mat &image_dst, int width_aperture) {/////////////////////////////redistribution image (if required)/////////////////////////////////////////////int Width_dst=image_src.cols;int height_Dst=image_Src.rows; Image_dst.create (Size (WIDTH_DST,HEIGHT_DST), CV_8UC1);//If reassigned, the previous space is discarded Image_dst.setto (Scalar (0));//Set to 0// sliding window int pixelcount=width_aperture*width_aperture;//The total number of pixels in the window Mat image_aperture (width_aperture,width_aperture,cv_ 8UC1)///Slide window image//histogram mat histogram;int histogramsize=256;//gray level int THRESHOLDVALUE=PIXELCOUNT/2+1;//STEP1. Set the Threshold value ( Step reference: High-efficiency programming points of the image four)//calculate the starting point coordinate int startx=width_aperture/2;int starty=width_aperture/2;/////////////////////////////// Expand the image boundary///////////////////////////////////////////int Height_extend=starty;int width_extend=startx; Mat Image_new;copymakeborder (image_src,image_new,height_extend,height_extend,width_extend,width_extend,border_ default);//Border_reflectint height_new=image_new.rows;int width_new=image_new.cols;//First line///Here you need to set 3 pointers, slide together/ 1. Pixels processed in the source image//2. MeshThe pixel//3 The marked image is processed. Source Image Sliding window Uchar *row_new=image_new.data+starty*width_new+startx;//New Image Uchar *row_dst=image_dst.data;// Target image Uchar *row_aperture_new=image_new.data;//the sliding window in the source image for (int y=starty;y<=height_new-starty-1;++y) {//Column Uchar * Col_new=row_new;uchar *col_dst=row_dst;uchar *col_aperture_new=row_aperture_new;//operation of the entire sliding window////////////////////// For sliding window operations/////////////////////////////////////////////calculates the first sliding window histogram per row//Extract sliding window image Uchar *row_aperture=image_ aperture.data;//sliding window image uchar *row_aperture_new_2=col_aperture_new;for (int k=0;k<=width_aperture-1;++k) {// Column Uchar *col_apertureimage=row_aperture;uchar *col_aperture_new_2=row_aperture_new_2;for (int w=0;w<=width_ APERTURE-1;++W) {//process each pixel col_apertureimage[0]=col_aperture_new_2[0];//next pixel col_apertureimage++;col_aperture_new _2++;} Next line Row_aperture+=width_aperture;row_aperture_new_2+=width_new;} Step 2. Determine the median value and record the number of pixels in the value of the luminance <=//Histogram calchist (&image_aperture, 1,//mat number 0,//used to calculate the histogram of the channel index, the channel index in order to row the mat (),// Mat () returns a null value, indicating no mask,histogram,//histogram 1,//StraightFanthou dimension, if 2 histogram is calculated, it is 2&histogramsize,//histogram rank number (such as gray level), that is, the number of rows per column 0//component variation range);//The number of pixels in the luminance median and <= value int Medianvalue , Pixlecountlowermedian; Calculateimage_mediangray_pixelcount (Histogram,pixelcount,medianvalue,pixlecountlowermedian);///////////////// Sliding window operation End//////////////////////////////////////////////filter col_dst[0]= (UCHAR) medianvalue;//slide one pixel ( Three pointers moving together) col_dst++;col_new++;col_aperture_new++;for (int x=startx+1;x<=width_new-startx-1;++x)//start with the second filter pixel per line {//////////////////////////////////Maintain sliding window histogram//////////////////////////////////////////step 3. Remove left Uchar *col_Left =col_aperture_new-1;float *data= (float*) histogram.data;for (int k=0;k<=width_aperture-1;++k) {int gray=col_Left[ 0];d ata[gray]-=1.0;if (gray<=medianvalue) {pixlecountlowermedian--;} Col_left+=width_new;} Step 4: Add right Uchar *col_right=col_aperture_new+width_aperture-1;for (int k=0;k<=width_aperture-1;++k) {int gray= Col_right[0];d ata[gray]+=1.0;if (gray<=medianvalue) {pixlecountlowermedian++;} Col_right+=width_new;} //Search for the new median if (Pixlecountlowermedian>thresholdvalue)//step 6. {while (1) {pixlecountlowermedian-=data[medianvalue];medianvalue--;if (pixlecountlowermedian<=thresholdvalue) { Break;}}} Else{while (pixlecountlowermedian<thresholdvalue)//step 5{medianvalue++;p ixlecountlowermedian+=data[ Medianvalue];}} Filter col_dst[0]= (UCHAR) medianvalue;//slide one pixel col_new++;col_dst++;col_aperture_new++;} End of x//next line row_new+=width_new;row_dst+=width_dst;row_aperture_new+=width_new;} End of Y}

Original:

High-efficiency median filtering (using Copymakeborder to process boundary pixels)