自适应中值滤波

作为代码界的菜鸟，最近在尝试着用vs实现《数字图像处理与机器视觉》一书中有关图像处理的VC++代码。目前在研究空间域增强，希望自己能够越走越远^_^
本篇是改进后的中值滤波即自适应中值滤波的实现。

#include <iostream>
#include <opencv2\core\core.hpp>
#include <opencv2\highgui\highgui.hpp>
#include <opencv2\imgproc\imgproc.hpp>
using namespace std;
using namespace cv;
/**************************************************
采用冒泡法对数组进行排序并返回数组元素中值
***************************************************/
int GetMedianValue(int *pAryGray, int nFilterLen)
{
    int i, j;
    int nMedianValue;
    int nTemp;
    for (j = 0; j < nFilterLen - 1; j++)
    {
        for (i = 0; i < nFilterLen - j - 1; i++)
        {
            if (pAryGray[i]>pAryGray[i + 1])
            {
                nTemp = pAryGray[i];
                pAryGray[i] = pAryGray[i + 1];
                pAryGray[i + 1] = nTemp;
            }
        }
    }
    if ((nFilterLen & 1) > 0)
    {
        nMedianValue = pAryGray[(nFilterLen + 1) / 2];
    }
    else
    {
        nMedianValue = (pAryGray[nFilterLen / 2] + pAryGray[nFilterLen / 2 + 1]) / 2;
    }
    return nMedianValue;
}
/********************************************************************************
中值滤波，对突发性噪声如椒盐噪声有较好的抑制效果
nFilterH：滤波器的高度，nFilterW：滤波器的宽度，
nFilterMX：滤波器的中心元素X坐标，nFilterMY：滤波器的中心元素Y坐标
*********************************************************************************/
void MedianFilter(Mat img, Mat &dst, int nFilterH, int nFilterW, int nFilterMY, int nFilterMX)
{
    int i, j, k, l;
    int nHeight = img.rows;
    int nWidth = img.cols;
    int *pAryR = new int[nFilterH*nFilterW];
    int *pAryG = new int[nFilterH*nFilterW];
    int *pAryB = new int[nFilterH*nFilterW];
    int ngR, ngG, ngB;
    for (i = nFilterMY; i < nHeight - nFilterH + nFilterMY + 1; i++)
    {
        for (j = nFilterMX; j < nWidth - nFilterW + nFilterMX + 1; j++)
        {
            for (k = 0; k < nFilterH; k++)
            {
                for (l = 0; l < nFilterW; l++)
                {
                    ngR = img.at<Vec3b>(i + 1 - nFilterMX, j + 1 - nFilterMY)[0];
                    pAryR[k*nFilterW + 1] = ngR;
                    ngG = img.at<Vec3b>(i + 1 - nFilterMX, j + 1 - nFilterMY)[1];
                    pAryG[k*nFilterW + 1] = ngG;
                    ngB = img.at<Vec3b>(i + 1 - nFilterMX, j + 1 - nFilterMY)[2];
                    pAryB[k*nFilterW + 1] = ngB;
                }
            }
    //对中值滤波进行改进，判断当前像素时候是领域的极大或极小值，如果是，将原图像像素值替换为中值，否则不改变原图像的值
            //如果只想看中值滤波的效果，可以将下面判断条件省去，直接将源图像像素值替换为中值即可
            ngR = GetMedianValue(pAryR, nFilterH*nFilterW);
            if (img.at<Vec3b>(i, j)[0] == pAryR[0] || img.at<Vec3b>(i, j)[0] == pAryR[nFilterH*nFilterW - 1])
                dst.at<Vec3b>(i, j)[0] = ngR;
            else
                dst.at<Vec3b>(i, j)[0] = img.at<Vec3b>(i, j)[0];


            ngG = GetMedianValue(pAryG, nFilterH*nFilterW);
            if (img.at<Vec3b>(i, j)[1] == pAryG[0] || img.at<Vec3b>(i, j)[1] == pAryG[nFilterH*nFilterW - 1])
                dst.at<Vec3b>(i, j)[1] = ngG;
            else
                dst.at<Vec3b>(i, j)[1] = img.at<Vec3b>(i, j)[1];


            ngB = GetMedianValue(pAryB, nFilterH*nFilterW);
            if (img.at<Vec3b>(i, j)[2] == pAryB[0] || img.at<Vec3b>(i, j)[2] == pAryB[nFilterH*nFilterW - 1])
                dst.at<Vec3b>(i, j)[2] = ngB;
            else
                dst.at<Vec3b>(i, j)[2] = img.at<Vec3b>(i, j)[2];
        }
    }
    delete[]pAryR;
    delete[]pAryG;
    delete[]pAryB;
}
void main()
{
    Mat img = imread("2.jpg");
    cvNamedWindow("原图", CV_WINDOW_FREERATIO);
    imshow("原图", img);
    Mat dst=img.clone();    
    中值滤波
    MedianFilter(img, dst, 7, 7, 1, 1);
    cvNamedWindow("中值滤波", CV_WINDOW_FREERATIO);
    imshow("中值滤波", dst);
    waitKey(0);
}

原图：

结果图：

自适应中值滤波对这个图像的作用不是特别明显，感兴趣的同学可以尝试用其他图片试验中值滤波的效果。，如果只想看中值滤波的效果，可以将中值滤波函数中的判断部分注释掉，直接替换中值即可