opencv 模糊C均值 c++

#include "all.h" double FCMeans(int *first,int* result, int clusterNum, int cols, int rows, int featureNum, int m_Value) { int sampleNum = cols * rows; double j_before = 0; double j_now = 0; int flag = 0;//隶属度是否为1的标识符。 //创建来记录隶属度值 //Mat u_account= Mat::zeros(sampleNum, clusterNum, CV_64FC1); vector<double>u_account; u_account.resize(sampleNum*clusterNum); //创建聚类中心,并随机初始化 vector<double>center; center.resize(clusterNum*featureNum); for (int i = 0; i < clusterNum; i++) { for (int j = 0; j < featureNum; j++) center[i*featureNum + j] = (255.0 / clusterNum) *(i + 1); } //创建距离，并初始化 vector<double>distance; distance.resize(sampleNum*clusterNum); for (int i = 0; i < sampleNum; i++)//遍历像素点的循环 { for (int k = 0; k < clusterNum; k++)//聚类循环 { double dis = 0; for (int n = 0; n < featureNum; n++) { double temp = 0; temp = center[k*featureNum + n] - first[i*featureNum + n]; dis = dis + temp *temp; } distance[i*clusterNum + k] = dis; } } j_before = 0; j_now = 0; //*********************************开始聚类 for (int i = 0; i < MAX; i++) { j_before = j_now;//记录上一次循环损失函数的值 j_now = 0; //初始化这一次损失函数的值 //计算隶属度矩阵 for (int i = 0; i < sampleNum; i++) { flag = 0; double top = 0;//每一个点的隶属度的分子 double base = 0;//每一个点的隶属度的分母；计算每个类的隶属度时分母不变 int k_save = 0;//记录距离为0的类别 for (int k = 0; k < clusterNum; k++) { double now_dis = 0;//记录当前当前坐标与聚类中心的距离 now_dis = distance[i*clusterNum + k]; if (now_dis != 0) { base = base + pow(1.0 / now_dis, 1.0 / (m_Value - 1));//算分母的值 } else { flag = 1; k_save = k; break; } } if (flag == 1) { for (int k = 0; k < clusterNum; k++) { u_account[i*clusterNum + k] = 0; } u_account[i*clusterNum + k_save] = 1; } else { for (int k = 0; k < clusterNum; k++) { top = pow(1.0 / distance[i*clusterNum + k], 1.0 / (m_Value - 1));//算分子的值 u_account[i*clusterNum + k] = top / base; } } } //修改新的聚类中心;聚类中心是featureNum个维度；clusterNum个;不同维度的分母值是一样的，分子不一样 for (int k = 0; k < clusterNum; k++)//类别遍历 { double base = 0; //每个维度的分母是一样的，所以在这里初始化，而且下面也只计算了一遍 for (int j = 0; j < featureNum; j++) { double top = 0; for (int i = 0; i < sampleNum; i++) { double a = 0;//分子分母的相似部分 a = pow(u_account[i*clusterNum + k], m_Value);///// if (j == 0) //只计算一次分母值 { base = base + a; } top = top + a * first[i*featureNum + j]; } center[k*featureNum + j] = top / base; } } //更新距离数组，同时计算j的值 for (int i = 0; i < sampleNum; i++)//遍历像素点的循环 { for (int k = 0; k < clusterNum; k++)//聚类循环 { double dis = 0; for (int n = 0; n < featureNum; n++) { double temp = 0; temp = center[k*featureNum + n] - first[i*featureNum + n]; dis = dis + temp *temp; } distance[i*clusterNum + k] = dis; j_now = j_now + pow(u_account[i*clusterNum + k], m_Value)*dis;//记录当前循环损失函数的值 } } if (fabsl(j_now - j_before) <= e) break; } ///聚类完毕，修改输出的像素值 for (int i = 0; i < sampleNum; i++) { int k_save = 0;//记录最大隶属度的类别 double temp = 0;//比较隶属度用的变量 for (int k = 0; k < clusterNum; k++) { if (u_account[i*clusterNum + k] > temp) { temp = u_account[i*clusterNum + k]; k_save = k; } } for (int j = 0; j < featureNum; j++) { result[i*featureNum + j] = center[k_save*featureNum + j]+0.5; } } /* for (int i = 0; i < 20; i++) { for (int j = 0; j < clusterNum; j++) { cout << u_account[i*clusterNum + j] << " "; } cout << endl; } //看隶属度值的函数 */ cout << endl; cout << clusterNum << "个聚类中心的像素值为: " << endl; for (int i = 0; i < clusterNum; i++) { for (int j = 0; j < featureNum; j++) { cout << center[i*featureNum + j] << " "; } cout << endl; }//看聚类中心的函数 return j_now; } void suit_FCM(int *first, int* result, int cols, int rows, int featureNum, int m_Value) { int num = cols*rows; double j_now = 0; double j_before = 0; j_before = FCMeans(first, result, 2, cols, rows, featureNum, 2); cout << j_before << endl; for (int i = 3; i < pow(num,0.5); i++)//pow(num, 0.5) { j_now = FCMeans(first, result, i, cols, rows, featureNum, 2); cout << j_now << endl; if (-j_now + j_before <= j_e) { cout << "最终聚类的类别为： " << i << endl; break; } j_before = j_now; } }