原理: 采用二维高斯模板对图像进行模糊处理，用于图像模糊化（去除细节和噪声），它的处理效果给人一种更佳柔和的感觉。

1.添加头文件, 并定义相关宏和结构体

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <jpeglib.h>
#include <math.h>

#define PI 3.141592654

#define JpegStdError jpeg_std_error
#define JpegCreateDecompress jpeg_create_decompress
#define JpegCreateCompress jpeg_create_compress
#define JpegStdioSrc jpeg_stdio_src
#define JpegReadHeader jpeg_read_header
#define JpegStartDecompress jpeg_start_decompress
#define ScanLine output_scanline
#define NextLine next_scanline
#define JpegReadLine jpeg_read_scanlines
#define JpegFinishDecompress jpeg_finish_decompress
#define JpegDestroyDecompress jpeg_destroy_decompress
#define JpegStdioDest jpeg_stdio_dest
#define JpegSetDefault jpeg_set_defaults
#define JpegSetQuality jpeg_set_quality
#define JpegStartCompress jpeg_start_compress
#define JpegWriteLine jpeg_write_scanlines
#define JpegFinishCompress jpeg_finish_compress
#define JpegDestroyCompress jpeg_destroy_compress

typedef JSAMPARRAY JpegArray;
typedef FILE *FilePtr;
typedef struct jpeg_compress_struct JpegComp;
typedef struct jpeg_decompress_struct JpegDecomp;
typedef struct jpeg_error_mgr JpegErrMgr;
typedef struct ImagePicData ImagePicData;
typedef struct ImageRGBAColor ImageRGBAColor;
typedef struct ImagePicPos ImagePicPos;
typedef struct ImageGaussBlurOpt ImageGaussBlurOpt;

struct ImagePicData
{
    unsigned char FName[256];
    FilePtr FPtr;
    int Width;
    int Height;
    int BitDepth;
    int Flag;
    int ColType;
    unsigned char *RGBA;
};

struct ImageRGBAColor
{
    int r;
    int g;
    int b;
    int a;
};

struct ImageGaussBlurOpt
{
    int Size;
    double Sigma;
    double *Kernel;
    unsigned char *Data;
};

2.创建高斯核

int CreateGaussKernel(ImageGaussBlurOpt *GaussBlur)
{
    double DistSqua;
    double Center = (GaussBlur->Size-1) / 2.0;
    double Sum = 0.0;
    double Value;
    int i, j;

    //生成高斯数据
    for(i=0; i<GaussBlur->Size; i++)
    {
        for(j=0; j<GaussBlur->Size; j++)
        {
             DistSqua = (i-Center)*(i-Center) + (j-Center)*(j-Center);
             Value = exp(-DistSqua/(2*GaussBlur->Sigma*GaussBlur->Sigma)) / (sqrt(2*PI)*GaussBlur->Sigma);
             GaussBlur->Kernel[i*GaussBlur->Size+j] = Value;
             Sum += Value;
        }
    }

    //归一化
    for(i=0; i<GaussBlur->Size; i++)
    {
        for(j=0; j<GaussBlur->Size; j++)
        {
            GaussBlur->Kernel[i*GaussBlur->Size+j] /= Sum;
            //printf("i=[%d], j=[%d], Kernel[i][j]=[%.5lf]\n", i, j, GaussBlur->Kernel[i*GaussBlur->Size+j]);
        }
    }

    return 0;
}

3.实现高斯滤波算法

int GaussBlurImage(ImagePicData *ImageData, ImageGaussBlurOpt *GaussBlur)
{
    ImagePicData ImageTempData;
    ImageRGBAColor RGBAColor;
    double *Kernel = GaussBlur->Kernel;
    int Border = (GaussBlur->Size-1) / 2;
    int Height = ImageData->Height;
    int Width = ImageData->Width;
    int Pos = 0;
    int i, j, k, r, s;

    ImageTempData.RGBA = GaussBlur->Data;

    for(i=Border; i<Height-Border; i++)
    {
        for(j=Border; j<Width-Border; j++)
        {
            RGBAColor.r = 0;
            RGBAColor.g = 0;
            RGBAColor.b = 0;

            for(r=-Border; r<=Border; r++)
            {
                for(s=-Border; s<=Border; s++)
                {
                    //printf("r=[%d], s=[%d], k=[%.5lf]\n", r, s, Kernel[(r+Border)*(GaussBlur->Size)+(s+Border)]);

                    RGBAColor.r += 1024*ImageData->RGBA[((i+r)*Width+(j+s))*4+0]*Kernel[(r+Border)*(GaussBlur->Size)+(s+Border)];
                    RGBAColor.g += 1024*ImageData->RGBA[((i+r)*Width+(j+s))*4+1]*Kernel[(r+Border)*(GaussBlur->Size)+(s+Border)];
                    RGBAColor.b += 1024*ImageData->RGBA[((i+r)*Width+(j+s))*4+2]*Kernel[(r+Border)*(GaussBlur->Size)+(s+Border)];
                }
            }

            //printf("i=[%d], j=[%d], r=[%d], g=[%d], b=[%d]\n", i, j, RGBAColor.r/1024, RGBAColor.g/1024, RGBAColor.b/1024);

            RGBAColor.r = (RGBAColor.r + 512) / 1024;
            RGBAColor.g = (RGBAColor.g + 512) / 1024;
            RGBAColor.b = (RGBAColor.b + 512) / 1024;

            if(RGBAColor.r < 0) RGBAColor.r = 0;
            else if(RGBAColor.r > 255) RGBAColor.r = 255;
            if(RGBAColor.g < 0) RGBAColor.g = 0;
            else if(RGBAColor.g > 255) RGBAColor.g = 255;
            if(RGBAColor.b < 0) RGBAColor.b = 0;
            else if(RGBAColor.b > 255) RGBAColor.b = 255;

            ImageTempData.RGBA[((i)*ImageData->Width+(j))*4+0] = RGBAColor.r;
            ImageTempData.RGBA[((i)*ImageData->Width+(j))*4+1] = RGBAColor.g;
            ImageTempData.RGBA[((i)*ImageData->Width+(j))*4+2] = RGBAColor.b;
        }
    }

    memcpy(ImageData->RGBA, ImageTempData.RGBA, Width*Height*4);

    return 0;
}

4.读写图片

参见《C语言实现色彩平衡算法》

5.添加主函数

int main(int argc, char **argv)
{
    ImagePicData PicData;
    ImageGaussBlurOpt GaussBlur;
    double Kernel[5*5];
    double Sigma = 0.01;
    char RGBAData[16*300*300];
    char RGBATempData[16*300*300];
    char JpegInputName[256];
    char JpegOutputName[256];

    memset(&PicData, 0x00, sizeof(PicData));
    memset(&GaussBlur, 0x00, sizeof(GaussBlur));
    memset(JpegInputName, 0x00, sizeof(JpegInputName));
    memset(JpegOutputName, 0x00, sizeof(JpegOutputName));

    if(argc != 4)
    {
        printf("入参个数错误!\n");
        return -1;
    }

    GaussBlur.Kernel = Kernel;
    GaussBlur.Data = RGBATempData;
    GaussBlur.Size = 5;
    GaussBlur.Sigma = 0.01;

    strncpy(JpegInputName, argv[1], 255);
    strncpy(JpegOutputName, argv[2], 255);

    PicData.RGBA = RGBAData;

    Sigma = atof(argv[3]);

    if(Sigma > 0.01) GaussBlur.Sigma = Sigma;

    strcpy(PicData.FName, JpegInputName);

    LoadJPG(&PicData);

    strcpy(PicData.FName, JpegOutputName);

    CreateGaussKernel(&GaussBlur);

    GaussBlurImage(&PicData, &GaussBlur);

    CreateJPG(&PicData);

    return 0;
}

6.编译运行

$ gcc -o examle examle.c -L$HOME/local/prior/lib -ljpeg -lm
$ ./examle a.jpg b.jpg 0.8

7.运行结果

原图

987cc94c21f05aa13c748e325ba49d10.png

效果

986cc96A.jpg