最小值滤波 (C 语言实现)

最小值滤波 (C 语言实现)

遇到最小值滤波的问题，小白不知道。一个程序写了三天,最终今天傍晚出来了。

。。

非常easy的for循环。可是没有理解最小值滤波。怎么写都是错啊~

这是我见过做好的描写叙述，关于最小值滤波：

3*3的像素点阵，对于中心点做最小值滤波的话，它的值将从77变换到0

处理结果图：

我一直支持也坚持开源分享的原则。为大家更好的相互学习，给出源码

/******************************************************************
code writer : EOF
code date : 2014.08.07
e-mail  : jasonleaster@gmail.com jasonleaster@163.com
 
code purpose:
	This demo is coded for mininum value filter.
If you find something wrong with my code, please touch me by e-mail.
Thank you.
 
*******************************************************************/
 
#include "opencv2/highgui/highgui_c.h"
#include "opencv2/imgproc/imgproc_c.h"
 
#include <stdio.h>
 
/*------------------------------------------------------------------------------
 
	This two Macro are used for debugging, if you are begginer with OpenCV,
it will help you to know and test what inside of the data struture in OpenCV
 
-------------------------------------------------------------------------------*/
 
//#define RGB_TEST_DEBUG
//#define CHANNEL_TEST_DEBUG	
 
/* the offset of three channel RGB */
#define RED_BIT   2
#define GREEN_BIT 1
#define BLUE_BIT  0
 
#define SQUARE_LENGTH 15
 
int get_dark_imagine(IplImage* const img_origin,IplImage* const img_win_dark);
 
int main(int argc,char* argv[])
{
	char* win_name_bf = "Before Processing";
	char* win_name_af = "After  Processing";
 
	CvSize size;
 
	IplImage* img_origin = cvLoadImage(argv[1],CV_LOAD_IMAGE_COLOR);
 
	size.height = img_origin->height;
	size.width  = img_origin->width;
 
	IplImage* img_win_dark = cvCreateImage(size,IPL_DEPTH_8U,1);//single channel
 
	get_dark_imagine(img_origin,img_win_dark);
 
	cvNamedWindow(win_name_bf,CV_WINDOW_AUTOSIZE);
	//cvNamedWindow is a function which would help you to creat a window.
 
	cvShowImage(win_name_bf,img_origin);
	//Obviously, show the picture that you inputed.
 
	cvNamedWindow(win_name_af,CV_WINDOW_AUTOSIZE);
	//cvNamedWindow is a function which would help you to creat a window.
 
	cvShowImage(win_name_af,img_win_dark);
	//Obviously, show the picture that you inputed.
 
	cvWaitKey(0);
	//pause and let the user see the picture.
 
	cvReleaseImage(&img_origin);
	cvReleaseImage(&img_win_dark);
	//Finally, release the struture, otherwise, memory leak !
 
	return 0;
}
 
int get_dark_imagine(IplImage* const img_origin,IplImage* const img_win_dark)
{
	/*
		Varible description:
 
			@img_origin  : A pointer which point to the original picture's IplImage-structure.
			@img_win_dark: A pointer which point to the dark-window's IplImage-structure.
	*/
 
	if(img_origin == NULL || img_win_dark == NULL)
	{
		printf("Error! img_origin or img_win_dark is NULL\n");
 
		return 1;
	}
 
	int height_origin = img_origin->height ;
	int width_origin  = img_origin->width ;//the search band width.
 
	unsigned char *  const ptr_array_origin    = (unsigned char*)img_origin->imageData;
	unsigned char *  const ptr_array_win_dark  = (unsigned char*)img_win_dark->imageData;
 
	unsigned char* ptr_header_origin    = NULL;
 
	int row = 0;
	int col = 0;
	int square_row = 0;
	int square_col = 0;
 
	int min    = 0;
	int T_min  = 0;
	int temp_R = 0;
	int temp_G = 0;
	int temp_B = 0;
	int temp   = 0;
 
	int search_win_start = SQUARE_LENGTH/2;
 
	/*
		 define two varible -- height_origin & width_origin for up band-width of the search-window
	*/
 
	int search_win_height_end  = img_win_dark->height - SQUARE_LENGTH/2;
	int search_win_width_end   = img_win_dark->width   - SQUARE_LENGTH/2;
 
	//initializition of the picture's data that 'ptr_array_win_dark' point to.
	for(row = 0; row < height_origin; row++)
	{
		for(col = 0; col < width_origin ;col++)
		{
			*(ptr_array_win_dark + col + row*(img_win_dark->widthStep)) = 255;
 
		}
	}
 
	//Mininum value filter
	for(row = search_win_start; row < search_win_height_end; row++)
	{
 
		for(col = search_win_start; col < search_win_width_end ;col++)
		{
 
			ptr_header_origin = ptr_array_origin + (row)*(img_origin->widthStep) + (col)*3;
 
			temp_B = *(ptr_header_origin + BLUE_BIT  );
			temp_G = *(ptr_header_origin + GREEN_BIT );
			temp_R = *(ptr_header_origin + RED_BIT   );
 
			min = (temp_G < temp_B) ?
 
temp_G :  temp_B;
			min = (min    < temp_R) ? min    :  temp_R;
 
			T_min = min;
 
			for(square_row = (row - search_win_start); square_row < (row + search_win_start + 1);square_row++)
			{
				for(square_col = (col - search_win_start); square_col < (col+search_win_start + 1);square_col++)
				{
					min = *(ptr_array_win_dark + square_col + square_row*(img_win_dark->widthStep));
 
					if (min > T_min)
					{
						*(ptr_array_win_dark + square_col + square_row*(img_win_dark->widthStep)) = (T_min);
					}
				}
			}
		}
	}
 
	return 0;
}

如有错误。欢迎交流指正

—— EOF

update : 2014.10.05

写了一个matlab版本号的最小滤波算法框架

Img_filted = dark_channel;
for row = 1 :  height
        for col = 1 : width            
 
                min_value =  dark_channel(row,col);
                for patch_row = (row  -floor(search_win_height/2)) : (row + floor(search_win_height/2))
                        for patch_col = (col - floor(search_win_width/2)) : (col   + floor(search_win_width/2))
 
                                if patch_row > 0  &&  patch_col > 0 && patch_row <= height && patch_col <= width
                                        if  min_value < Img_filted(patch_row,patch_col)
                                               Img_filted(patch_row,patch_col) = min_value;
                                       end
                                end
                        end
                end
        end
end