图像直方图均衡化(Equalization)数学原理和纯C语言实现

图像处理入门

C语言生成RGB24格式图像

图像缩放之最近邻内插法

图像直方图均衡化数学原理和程序实现

---

目录

图像处理入门

一、图像均衡化数学原理

1.转化为数学问题

2.需要用到的知识

3.数学推理

二、程序实现

1.程序实现原理

2.代码实现

3.程序执行效果

4.图像均衡化前后效果

总结

---

一、图像均衡化数学原理

1.转化为数学问题

        已知随机变量X的分布函数Fx(𝑥)和概率密度fx (x)，以及对随机变量进行Y=g(X)变化后的随机变量的概率密度1/(L-1)，g(x)为单调连续，求转换函数y=g(x)。

2.需要用到的知识

        一维函数微积分
        随机变量的概率分布

3.数学推理

  

二、程序实现

1.程序实现原理

2.代码实现

代码如下（示例）：

#include <stdlib.h>
#include <stdio.h>
 
 
#define PICT_WIDTH  90
#define PICT_HEIGHT  60
 
 
static unsigned int s_rgb[PICT_WIDTH][PICT_HEIGHT][3] = { 0 };
 
/**
* rgb24
* r:203 g:64,173,38 b:142
*/
int make_rgb24_colorbar(unsigned int rgb[90][60][3])
{
	int i = 0;
	int j = 0;
	unsigned char chr = 0;
 
	for (j = 0; j<PICT_HEIGHT; j++)
	{
		for (i = 0; i < PICT_WIDTH; i++)
		{
			if (i<30) 
			{
				chr = 203;
			}
			else if (i<60) 
			{
				chr = 64;
			}
			else 
			{
				chr = 0;
			}
			rgb[i][j][0] = chr;
 
			if (i<30)
			{
				chr = 0;
			}
			else if (i<60)
			{
				chr = 173;
			}
			else
			{
				chr = 0;
			}
			rgb[i][j][1] = chr;
 
			if (i<30)
			{
				chr = 0;
			}
			else if (i<60)
			{
				chr = 38;
			}
			else
			{
				chr = 142;
			}
			rgb[i][j][2] = chr;
		}
	}
	return 0;
}
 
int generate_rgb24_file(unsigned int rgb[90][60][3],char *file_name)
{
	char file_path[64] = {0};
	if (file_name == NULL) 
	{
		return 0;
	}
	snprintf(file_path,63,"rgb_%s_90_60.rgb24", file_name);
	FILE *fp = fopen(file_path, "wb+");
	int i = 0;
	int j = 0;
	unsigned char chr = 0;
 
	for (j = 0; j< PICT_HEIGHT; j++)
	{
		for (i = 0; i < PICT_WIDTH; i++)
		{
			chr = rgb[i][j][0];
			fwrite(&chr, 1, 1, fp);
 
			chr = rgb[i][j][1];
			fwrite(&chr, 1, 1, fp);
 
			chr = rgb[i][j][2];
			fwrite(&chr, 1, 1, fp);
		}
	}
	fclose(fp);
	return 0;
}
 
#define IMAGE_COLOR_LEVEL 256
 
#define HISTOGRAM_HEIGHT_MAX 20
#define HISTOGRAM_WIDTH_MAX 128
 
#define STATISTIC_DEC_MAX 5
#define STATISTIC_HOR_AXS_NUM (HISTOGRAM_WIDTH_MAX / STATISTIC_DEC_MAX)
 
int print_histogram(int image_statistic[IMAGE_COLOR_LEVEL])
{
	int *image_point = image_statistic;
	char hist_char = '*';
	char hist_space_char = ' ';
	char disp_array[HISTOGRAM_HEIGHT_MAX][HISTOGRAM_WIDTH_MAX + 1] = {0};
	int hist_statistic[HISTOGRAM_WIDTH_MAX] = {0};
	int i = 0;
	int j = 0;
	int k = 0;
	int statistic_max = 0;
	int hist_hor_axis[STATISTIC_HOR_AXS_NUM] = { 0 };
 
	for (i = 0; i < IMAGE_COLOR_LEVEL; i++)
	{
		hist_statistic[i * HISTOGRAM_WIDTH_MAX / IMAGE_COLOR_LEVEL] += image_statistic[i];
	}
 
	for (i = 0; i < STATISTIC_HOR_AXS_NUM; i++)
	{
		hist_hor_axis[i] = 1.0 * STATISTIC_DEC_MAX * i * IMAGE_COLOR_LEVEL / HISTOGRAM_WIDTH_MAX;
	}
 
	for (i = 0; i < HISTOGRAM_WIDTH_MAX; i++)
	{
		if (hist_statistic[i] > statistic_max)
		{
			statistic_max = hist_statistic[i];
		}
	}
 
 
	for (i = 0; i < HISTOGRAM_WIDTH_MAX;i++)
	{
		int hist_height = hist_statistic[i] * HISTOGRAM_HEIGHT_MAX / statistic_max;
		for (j = 0; j < HISTOGRAM_HEIGHT_MAX; j++)
		{
			if (j < hist_height)
			{
				disp_array[j][i] = hist_char;
			}
			else
			{
				disp_array[j][i] = hist_space_char;
			}
		}
	}
 
	printf(
		"     y\n"
		"      \n"
		"     |\n"
		"%5d|\n", statistic_max);
	for (i = 0; i < HISTOGRAM_HEIGHT_MAX; i++)
	{
		printf(
		"     |%s\n", disp_array[HISTOGRAM_HEIGHT_MAX - i - 1]);
	}
	printf(
		"     |--------------------------------------------------------------------------------------------------------------------------------->x\n");
	printf(
		"     ");
	char dec_max[32] = { 0 };
	char format_str[32] = "%-";
	sprintf(dec_max, "%d", STATISTIC_DEC_MAX);
	strncat(format_str, dec_max,32);
	strncat(format_str, "d", 32);
	for (i = 0; i < STATISTIC_HOR_AXS_NUM; i++)
	{
		printf(
			format_str, hist_hor_axis[i]);
	}
	printf(
		"\n");
	return 0;
}
 
int statistic_histogram(unsigned int rgb[90][60][3],int image_statistic[IMAGE_COLOR_LEVEL])
{
	int i = 0;
	int j = 0;
	int k = 0;
	for(i = 0; i < 90; i++)
	{
		for (j = 0; j < 60; j++) 
		{
			for (k = 0; k < 3; k++)
			{
				image_statistic[rgb[i][j][k]]++;
			}
		}
	}
	return 0;
}
 
//直方图均衡化
int histogram_equalization(unsigned int rgb[90][60][3], unsigned int result_rgb[90][60][3])
{
	int image_statistic[IMAGE_COLOR_LEVEL] = {0};
	int image_statistic2[IMAGE_COLOR_LEVEL] = {0};
	int i = 0;
	int j = 0;
	int k = 0;
	int tmp = 0;
	int pixel_num = 90 * 60 * 3;
	float ratio = 0;
	float equal_ratio[IMAGE_COLOR_LEVEL] = { 0 };
 
	statistic_histogram(rgb, image_statistic);
	
	for (i = 0; i < IMAGE_COLOR_LEVEL; i++)
	{
		if(image_statistic[i] != 0)
		{
			printf("before,i:[%d],num:[%d]\n",i,image_statistic[i]);
		}
	}
	
	for (i = 0; i < IMAGE_COLOR_LEVEL; i++)
	{
		for (j = 0; j < i + 1; j++) 
		{
			equal_ratio[i] += image_statistic[j] * 1.0 / pixel_num;
		}
	}
 
	for (i = 0; i < 90; i++)
	{
		for (j = 0; j < 60; j++)
		{
			for (k = 0; k < 3; k++)
			{
				result_rgb[i][j][k] = (unsigned int)(IMAGE_COLOR_LEVEL - 1) * equal_ratio[rgb[i][j][k]];
			}
		}
	}
 
	statistic_histogram(result_rgb, image_statistic2);
	
	for (i = 0; i < IMAGE_COLOR_LEVEL; i++)
	{
		if(image_statistic2[i] != 0)
		{
			printf("after,i:[%d],num:[%d]\n",i,image_statistic2[i]);
		}
	}
	
	return 0;
}

int main()
{
	int image_statistic[IMAGE_COLOR_LEVEL] = {0};
	make_rgb24_colorbar(s_rgb);
	generate_rgb24_file(s_rgb, "colorbar");
	statistic_histogram(s_rgb, image_statistic);
	print_histogram(image_statistic);
 
	unsigned int result_rgb[90][60][3];
	int result_statistic[IMAGE_COLOR_LEVEL] = { 0 };
	histogram_equalization(s_rgb, result_rgb);
	statistic_histogram(result_rgb, result_statistic);
	print_histogram(result_statistic);
	generate_rgb24_file(result_rgb,"equalization");
	sleep(100);;
    return 0;
}
 

3.程序执行效果

4.图像均衡化前后效果

均衡化前：

均衡化后：

 

---

总结

均衡化后直方图更均匀，但是实际图像质量不一定更好。