Opencv中Kmeans使用学习，更换证件照背景，QT界面，函数实现，C++实现

kmeans简介

kmeans是机器学习中的一种聚类算法，简单来说就是把在一个区域中的点进行分类，分类的类别由自己定。具体的数学原理大家可以在很多的blog中学习，在这里主要介绍opencv中python的实现。在这里给大家一个非常直观的链接https://www.naftaliharris.com/blog/visualizing-k-means-clustering/大家可以访问该网址。里面由DBSCAN,kmeans等算法的可视化显示，下面给出一定的截图
通过这个可视化模型，大家可以非常好的学习kmeans和DBSCAN两种聚类算法

python代码实现

#__aditor:dwy
#__date: 2019/5/29
#!/usr/bin/env python    
# -*- coding: utf-8 -*-
import cv2 as cv
import numpy as np
img = cv.imread("tx.png")

#准备数据
color=[0,0,255]
wide = img.shape[1]
height = img.shape[0]
dims =img.shape[2]
sampleCount = wide*height
clusterCount = 4
points = np.zeros([sampleCount,dims],np.float32)#把准备好的数据存储到points中
centers = np.zeros([clusterCount,1],np.float32)
index = 0
for i in range(height):
	for j in range(wide):
		index = i*wide+j
		for n in range(dims):
			points[index][n]=img[i][j][n]
criteria = (cv.TERM_CRITERIA_EPS + cv.TERM_CRITERIA_MAX_ITER, 10, 0.1)
ret,label,center=cv.kmeans(points,clusterCount,None,criteria,10,cv.KMEANS_PP_CENTERS)#聚类的结果在ret中，标签为label ，中心为center
mask = np.zeros((img.shape[0],img.shape[1]),np.uint8)#创建一个掩模，如果是背景则为黑，不为背景则为255
index = wide*2+4#把第二行的第四个数据作为背景的标记（通常这个区域一般为背景，也可以设置某类包含点最多的为背景）
img2=img.copy()
cindex = label[index]#创建掩模的过程
for i in range(height):
	for j in range(wide):
		index = i * wide + j
		label1 = label[index]
		if cindex==label1:
			for n in range(dims):
				img2[i][j][n]=0;
			mask[i][j]=0;
		else:
			mask[i][j]=255
structuting = cv.getStructuringElement(cv.MORPH_RECT,(11,11))#对图像进行开操作和膨胀处理
mask=cv.erode(mask,structuting)
mask=cv.GaussianBlur(mask,(9,9),0,0)#下面的循环是按照掩模复制图像，如果为0则设置成要替换的背景，如果不为0则把原图中的像素填充进去，clolr为预定的颜色
for i in range(height):
	for j in range(wide):
		if mask[i][j]==0:
			for n in range(dims):
				img2[i][j][n]=color[n]
		else:
			continue
cv.imshow("mask",img2)
cv.imshow("img",img)

cv.waitKey(0)

下面是代码运行的结果
左边是更换前的代码，右边是更换后的代码。
为了更好的使用该方法，本文提供qt可视化的更改代码

# -*- coding: utf-8 -*-

# Form implementation generated from reading ui file 'D:\My code\Python\PYQT\Background\background.ui'
#
# Created by: PyQt5 UI code generator 5.11.3
#
# WARNING! All changes made in this file will be lost!

from PyQt5 import QtCore, QtGui, QtWidgets

class Ui_MainWindow(object):
    def setupUi(self, MainWindow):
        MainWindow.setObjectName("MainWindow")
        MainWindow.resize(800, 600)
        self.centralWidget = QtWidgets.QWidget(MainWindow)
        self.centralWidget.setObjectName("centralWidget")
        self.gridLayout_4 = QtWidgets.QGridLayout(self.centralWidget)
        self.gridLayout_4.setObjectName("gridLayout_4")
        self.widget_2 = QtWidgets.QWidget(self.centralWidget)
        self.widget_2.setObjectName("widget_2")
        self.gridLayout_2 = QtWidgets.QGridLayout(self.widget_2)
        self.gridLayout_2.setObjectName("gridLayout_2")
        self.label_2 = QtWidgets.QLabel(self.widget_2)
        self.label_2.setObjectName("label_2")
        self.gridLayout_2.addWidget(self.label_2, 0, 0, 1, 1)
        self.label_4 = QtWidgets.QLabel(self.widget_2)
        self.label_4.setObjectName("label_4")
        self.gridLayout_2.addWidget(self.label_4, 0, 1, 1, 1)
        self.before = QtWidgets.QLabel(self.widget_2)
        self.before.setText("")
        self.before.setObjectName("before")
        self.gridLayout_2.addWidget(self.before, 1, 0, 1, 1)
        self.after = QtWidgets.QLabel(self.widget_2)
        self.after.setText("")
        self.after.setObjectName("after")
        self.gridLayout_2.addWidget(self.after, 1, 1, 1, 1)
        self.gridLayout_2.setRowStretch(0, 1)
        self.gridLayout_2.setRowStretch(1, 9)
        self.gridLayout_4.addWidget(self.widget_2, 0, 0, 1, 1)
        self.widget_3 = QtWidgets.QWidget(self.centralWidget)
        self.widget_3.setObjectName("widget_3")
        self.horizontalLayout = QtWidgets.QHBoxLayout(self.widget_3)
        self.horizontalLayout.setObjectName("horizontalLayout")
        spacerItem = QtWidgets.QSpacerItem(148, 20, QtWidgets.QSizePolicy.Expanding, QtWidgets.QSizePolicy.Minimum)
        self.horizontalLayout.addItem(spacerItem)
        self.widget = QtWidgets.QWidget(self.widget_3)
        self.widget.setObjectName("widget")
        self.gridLayout = QtWidgets.QGridLayout(self.widget)
        self.gridLayout.setObjectName("gridLayout")
        self.loadImage = QtWidgets.QPushButton(self.widget)
        self.loadImage.setObjectName("loadImage")
        self.gridLayout.addWidget(self.loadImage, 0, 0, 1, 1)
        self.colorchoose = QtWidgets.QComboBox(self.widget)
        self.colorchoose.setObjectName("colorchoose")
        self.colorchoose.addItem("")
        self.colorchoose.addItem("")
        self.colorchoose.addItem("")
        self.colorchoose.addItem("")
        self.colorchoose.addItem("")
        self.colorchoose.addItem("")
        self.colorchoose.addItem("")
        self.gridLayout.addWidget(self.colorchoose, 0, 1, 1, 1)
        self.change = QtWidgets.QPushButton(self.widget)
        self.change.setObjectName("change")
        self.gridLayout.addWidget(self.change, 0, 2, 1, 1)
        self.pushButton_2 = QtWidgets.QPushButton(self.widget)
        self.pushButton_2.setObjectName("pushButton_2")
        self.gridLayout.addWidget(self.pushButton_2, 1, 0, 1, 1)
        self.pushButton_4 = QtWidgets.QPushButton(self.widget)
        self.pushButton_4.setObjectName("pushButton_4")
        self.gridLayout.addWidget(self.pushButton_4, 1, 1, 1, 1)
        self.pushButton_5 = QtWidgets.QPushButton(self.widget)
        self.pushButton_5.setObjectName("pushButton_5")
        self.gridLayout.addWidget(self.pushButton_5, 1, 2, 1, 1)
        self.horizontalLayout.addWidget(self.widget)
        self.groupBox = QtWidgets.QGroupBox(self.widget_3)
        self.groupBox.setObjectName("groupBox")
        self.gridLayout_3 = QtWidgets.QGridLayout(self.groupBox)
        self.gridLayout_3.setObjectName("gridLayout_3")
        self.label_5 = QtWidgets.QLabel(self.groupBox)
        self.label_5.setObjectName("label_5")
        self.gridLayout_3.addWidget(self.label_5, 0, 0, 1, 1)
        self.red = QtWidgets.QLineEdit(self.groupBox)
        self.red.setText("")
        self.red.setObjectName("red")
        self.gridLayout_3.addWidget(self.red, 0, 1, 1, 1)
        self.label_7 = QtWidgets.QLabel(self.groupBox)
        self.label_7.setObjectName("label_7")
        self.gridLayout_3.addWidget(self.label_7, 1, 0, 1, 1)
        self.green = QtWidgets.QLineEdit(self.groupBox)
        self.green.setText("")
        self.green.setObjectName("green")
        self.gridLayout_3.addWidget(self.green, 1, 1, 1, 1)
        self.label_6 = QtWidgets.QLabel(self.groupBox)
        self.label_6.setObjectName("label_6")
        self.gridLayout_3.addWidget(self.label_6, 2, 0, 1, 1)
        self.blue = QtWidgets.QLineEdit(self.groupBox)
        self.blue.setText("")
        self.blue.setObjectName("blue")
        self.gridLayout_3.addWidget(self.blue, 2, 1, 1, 1)
        self.gridLayout_3.setColumnStretch(0, 1)
        self.gridLayout_3.setColumnStretch(1, 2)
        self.horizontalLayout.addWidget(self.groupBox)
        spacerItem1 = QtWidgets.QSpacerItem(148, 20, QtWidgets.QSizePolicy.Expanding, QtWidgets.QSizePolicy.Minimum)
        self.horizontalLayout.addItem(spacerItem1)
        self.gridLayout_4.addWidget(self.widget_3, 1, 0, 1, 1)
        self.textBrowser = QtWidgets.QTextBrowser(self.centralWidget)
        font = QtGui.QFont()
        font.setFamily("宋体")
        font.setPointSize(12)
        self.textBrowser.setFont(font)
        self.textBrowser.setObjectName("textBrowser")
        self.gridLayout_4.addWidget(self.textBrowser, 2, 0, 1, 1)
        self.gridLayout_4.setRowStretch(0, 3)
        self.gridLayout_4.setRowStretch(1, 1)
        self.gridLayout_4.setRowStretch(2, 1)
        MainWindow.setCentralWidget(self.centralWidget)

        self.retranslateUi(MainWindow)
        QtCore.QMetaObject.connectSlotsByName(MainWindow)

    def retranslateUi(self, MainWindow):
        _translate = QtCore.QCoreApplication.translate
        MainWindow.setWindowTitle(_translate("MainWindow", "MainWindow"))
        self.label_2.setText(_translate("MainWindow", "处理前图像"))
        self.label_4.setText(_translate("MainWindow", "处理后图像"))
        self.loadImage.setText(_translate("MainWindow", "导入图片"))
        self.colorchoose.setItemText(0, _translate("MainWindow", "蓝色"))
        self.colorchoose.setItemText(1, _translate("MainWindow", "白色"))
        self.colorchoose.setItemText(2, _translate("MainWindow", "红色"))
        self.colorchoose.setItemText(3, _translate("MainWindow", "灰色"))
        self.colorchoose.setItemText(4, _translate("MainWindow", "紫色"))
        self.colorchoose.setItemText(5, _translate("MainWindow", "黄色"))
        self.colorchoose.setItemText(6, _translate("MainWindow", "绿色"))
        self.change.setText(_translate("MainWindow", "确认更改"))
        self.pushButton_2.setText(_translate("MainWindow", "待添加功能"))
        self.pushButton_4.setText(_translate("MainWindow", "待添加功能"))
        self.pushButton_5.setText(_translate("MainWindow", "待添加功能"))
        self.groupBox.setTitle(_translate("MainWindow", "RGB通道"))
        self.label_5.setText(_translate("MainWindow", "    R"))
        self.label_7.setText(_translate("MainWindow", "    G"))
        self.label_6.setText(_translate("MainWindow", "    B"))
        self.textBrowser.setHtml(_translate("MainWindow", "<!DOCTYPE HTML PUBLIC \"-//W3C//DTD HTML 4.0//EN\" \"http://www.w3.org/TR/REC-html40/strict.dtd\">\n"
"<html><head><meta name=\"qrichtext\" content=\"1\" /><style type=\"text/css\">\n"
"p, li { white-space: pre-wrap; }\n"
"</style></head><body style=\" font-family:\'宋体\'; font-size:12pt; font-weight:400; font-style:normal;\">\n"
"<p style=\" margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px; -qt-block-indent:0; text-indent:0px;\"><span style=\" font-family:\'SimSun\';\">以后还有什么图像处理相关的内容可以在此基础上扩展，RGB通道可以更改到任意背景颜色</span></p></body></html>"))


if __name__ == "__main__":
    import sys
    app = QtWidgets.QApplication(sys.argv)
    MainWindow = QtWidgets.QMainWindow()
    ui = Ui_MainWindow()
    ui.setupUi(MainWindow)
    MainWindow.show()
    sys.exit(app.exec_())

界面布局代码部分文件命名为Ui_background.py为后一个文件导入方便

# -*- coding: utf-8 -*-

"""
Module implementing MainWindow.
"""

from PyQt5.QtCore import *
from PyQt5.QtGui import *
from PyQt5 import QtWidgets
from PyQt5.QtWidgets import *
from Ui_background import Ui_MainWindow   #如果此处的ui文件命名不同请改命名
import cv2 as cv
import numpy as np
import os


class MainWindow(QMainWindow, Ui_MainWindow):
    """
    Class documentation goes here.
    """
    def __init__(self, parent=None):
        """
        Constructor
        
        @param parent reference to the parent widget
        @type QWidget
        """
        super(MainWindow, self).__init__(parent)
        self.setupUi(self)
        self.color = [0, 0, 0]
        self.target =True 
        self.i=0
        self.newname =''
        self.file_name=""
        self.file_type =""
    
    @pyqtSlot()
    def on_loadImage_clicked(self):
        """
        Slot documentation goes here.
        """
        # TODO: not implemented yet
        
        self.file_name, self.file_type = QFileDialog.getOpenFileName(self,"打开文件","D:/","All Files (*);")

        ##"open file Dialog "为文件对话框的标题，第三个是打开的默认路径，第四个是文件类型过滤器
        img = cv.imread(self.file_name, -1)


        if (img.shape[2] == 3):#bgr到rgb转换的显示
            qimg = cv.cvtColor(img, cv.COLOR_BGR2RGB)
            qimg = QImage(qimg.data, qimg.shape[1], qimg.shape[0], qimg.shape[1]*qimg.shape[2],QImage.Format_RGB888)
            self.before.setPixmap(QPixmap.fromImage(qimg))
        else:
            qimg = QImage(img.data, img.shape[1], img.shape[0],QImage.Format_Grayscale8)
            self.before.setPixmap(QPixmap.fromImage(qimg))
    
        

    
    @pyqtSlot(str)
    def on_colorchoose_currentIndexChanged(self, p0):
        """
        Slot documentation goes here.
        
        @param p0 DESCRIPTION
        @type str
        """
        # TODO: not implemented yet
        current = self.colorchoose.currentText()
        if self.target:
            if current=="红色":
                self.color =[0, 0, 255]
                self.newname= '_red'
            elif current=="白色":
                self.color =[255, 255, 255]
                self.newname= '_white'
            elif current=="蓝色":
                self.color =[255, 0, 0]
                self.newname= '_blue'
            elif current=="绿色":
                self.color =[0, 255, 0]
                self.newname= '_green'
            elif current=="黄色":
                self.color=[0, 255, 255]
                self.newname= '_yellow'
            elif current=="紫色":
                self.color=[255, 0, 255]
                self.newname= '_violet'
            elif current=="灰色":
                self.color=[96, 96, 96]
                self.newname= '_gray'
    
    @pyqtSlot()
    def on_change_clicked(self):
        """
        Slot documentation goes here.
        """
        # TODO: not implemented yet
        if self.red.text()!="":
            self.red1 =int(self.red.text())
            self.green1 =int(self.green.text())
            self.blue1 =int(self.blue.text())
            self.color = [self.blue1, self.green1,self.red1]
            self.newname='_define'
            print(self.color)
            self.target = False
        img =cv.imread(self.file_name, -1)

        if (img is None):
            print(1)
            self.textBrowser.setText("打开图片失败，请重新加载")
        else:
            # 准备数据
            wide = img.shape[1]
            height = img.shape[0]
            dims = img.shape[2]
            sampleCount = wide * height
            clusterCount = 4
            points = np.zeros([sampleCount, dims], np.float32)
            centers = np.zeros([clusterCount, 1], np.float32)
            index = 0
            for i in range(height):
                for j in range(wide):
                    index = i * wide + j
                    for n in range(dims):
                        points[index][n] = img[i][j][n]
            criteria = (cv.TERM_CRITERIA_EPS + cv.TERM_CRITERIA_MAX_ITER, 10, 0.1)
            ret, label, center = cv.kmeans(points, clusterCount, None, criteria, 10, cv.KMEANS_PP_CENTERS)
            mask = np.zeros((img.shape[0], img.shape[1]), np.uint8)
            index = wide * 2 + 4
            img2 = img.copy()
            cindex = label[index]
            for i in range(height):
                for j in range(wide):
                    index = i * wide + j
                    label1 = label[index]
                    if cindex == label1:
                        for n in range(dims):
                            img2[i][j][n] = 0;
                        mask[i][j] = 0;
                    else:
                        mask[i][j] = 255
            structuting = cv.getStructuringElement(cv.MORPH_RECT, (13, 13))
            mask =cv.dilate(mask,structuting)
            mask = cv.erode(mask, structuting)
            structuting2 = cv.getStructuringElement(cv.MORPH_RECT, (9, 9))
            mask = cv.erode(mask, structuting2)
            cv.waitKey(10)
            mask = cv.GaussianBlur(mask, (9, 9), 0, 0)
            for i in range(height):
                for j in range(wide):
                    if mask[i][j] == 0:
                        for n in range(dims):
                            img2[i][j][n] =  self.color[n]
                    else:
                        continue
            img2 = cv.GaussianBlur(img2, (3, 3), 0, 0)
            if (img2.shape[2] == 3):
                qimg = cv.cvtColor(img2, cv.COLOR_BGR2RGB)
                qimg = QImage(qimg.data, qimg.shape[1], qimg.shape[0], qimg.shape[1] * qimg.shape[2],
                              QImage.Format_RGB888)
                self.after.setPixmap(QPixmap.fromImage(qimg))
            else:
                qimg = QImage(img.data, img.shape[1], img.shape[0], QImage.Format_Grayscale8)
                self.after.setPixmap(QPixmap.fromImage(qimg))
            (fathername, filename) = os.path.split(self.file_name)#读取文件的父目录，文件名
            (name, type) = os.path.splitext(filename)#分开文件名后缀名
            name=fathername+"/"+name+self.newname+type
            print(name)
            cv.imwrite(name,img2)
            write = "the image has been saved in:"+name
            self.textBrowser.setText(write)

        
if __name__ == "__main__":
    import sys
    app = QtWidgets.QApplication(sys.argv)
    ui = MainWindow()
    ui.show()
    sys.exit(app.exec_())
    

实现部分
最后的实现效果如下图
只需要更改下拉框的参数便可以实现底色的改变。如果有需要的也可以用pyinstall 去打包py文件成可以执行的exe文件。
如果需要打包好的文件可以在下方留言，看到的话可以发给你

c++实现

int kmeans_picture()
{
	Mat src = imread("tx.png");
	Mat img2 = src.clone();
	if (src.empty()) {
		printf("could not load image...\n");
		return -1;
	}
	namedWindow("input image", WINDOW_AUTOSIZE);
	imshow("input image", src);

	Scalar colorTab[] = {
		Scalar(0, 0, 255),
		Scalar(0, 255, 0),
		Scalar(255, 0, 0),
		Scalar(0, 255, 255),
		Scalar(255, 0, 255)
	};
	int width = src.cols;
	int height = src.rows;
	int dims = src.channels();//获取图像的要素
    // 初始化定义，准备数据阶段，把数据展开成width *height行的数据
	int sampleCount = width*height;
	int clusterCount = 4;//分类的组别
	Mat points(sampleCount, dims, CV_32F, Scalar(10));//重新生成分类的数组
	Mat labels;
	Mat centers(clusterCount, 1, points.type());

	// RGB 数据转换到样本数据
	//int index = 0;
	//for (int row = 0; row < height; row++) {
	//	for (int col = 0; col < width; col++) {
	//		index = row*width + col;
	//		Vec3b bgr = src.at<Vec3b>(row, col);
	//		points.at<float>(index, 0) = static_cast<int>(bgr[0]);
	//		points.at<float>(index, 1) = static_cast<int>(bgr[1]);
	//		points.at<float>(index, 2) = static_cast<int>(bgr[2]);
	//	}
	//}
	//指针索引方法
	int index = 0;
	for (int row = 0; row < height; row++)
	{
		
		for (int col = 0; col < width; col++)
		{
			index = row*width + col;
			uchar *p = src.ptr<uchar>(row,col);
			points.at<float>(index, 0) = static_cast<int>(p[0]);
			points.at<float>(index, 1) = static_cast<int>(p[1]);
			points.at<float>(index, 2) = static_cast<int>(p[2]);
		}
	}

	// 运行K-Means
	TermCriteria criteria = TermCriteria(TermCriteria::EPS + TermCriteria::COUNT, 10, 0.1);
	kmeans(points, clusterCount, labels, criteria, 3, KMEANS_PP_CENTERS, centers);

	Mat mask = Mat::zeros(src.size(), CV_8UC1);
	index = src.rows * 2 + 2;
	int cindex = labels.at<int>(index, 0);//背景设置为0
	height = src.rows;
	width = src.cols;

	for (int row = 0;row<height;row++)
	{
		for (int col = 0;col<width;col++)
		{
			index = row*width + col;
			int label = labels.at<int>(index, 0);
			if (label == cindex)
			{
				img2.at<Vec3b>(row, col)[0] = 0;
				img2.at<Vec3b>(row, col)[1] = 0;
				img2.at<Vec3b>(row, col)[2] = 0;
				mask.at<uchar>(row, col) = 0;
			}
			else
			{
				mask.at<uchar>(row, col) = 255;
			}
		}
	}

	//腐蚀
	Mat k = getStructuringElement(MORPH_RECT, Size(3, 3), Point(-1, -1));
	erode(mask, mask, k);
	imshow("win4", mask);

	//高斯模糊
	GaussianBlur(mask, mask, Size(3, 3), 0, 0);
	imshow("win5", mask);

	//通道混合
	RNG rng(12345);
	Vec3b color;
	color[0] = rng.uniform(0, 255);
	color[1] = rng.uniform(0, 255);
	color[2] = rng.uniform(0, 255);
	Mat result(src.size(), src.type());

	double d1 = 0.0;
	int r = 0, g = 0, b = 0;
	int r1 = 0, g1 = 0, b1 = 0;
	int r2 = 0, g2 = 0, b2 = 0;


	for (int row = 0; row < height;row++)
	{
		for (int col = 0;col<width;col++)
		{
			int m = mask.at<uchar>(row, col);
			if (m == 255)
			{
				result.at<Vec3b>(row, col) = src.at<Vec3b>(row, col);//前景
			}
			else if (m == 0)
			{
				result.at<Vec3b>(row, col) = color;//背景
			}
			else
			{
				d1 = m / 255.0;
				b1 = src.at<Vec3b>(row, col)[0];
				g1 = src.at<Vec3b>(row, col)[1];
				r1 =src.at<Vec3b>(row, col)[2];

				b2 = color[0];
				g2 = color[1];
				r2 = color[2];

				b = b1*d1 + b2*(1.0 - d1);
				g = g1*d1 + g2*(1.0 - d1);
				r = r1*d1 + r2*(1.0 - d1);

				result.at<Vec3b>(row, col)[0] = b;
				result.at<Vec3b>(row, col)[1] = g;
				result.at<Vec3b>(row, col)[2] = r;
			}
		}
	}

	//输出
	imshow("win2", mask);
	imshow("win3", img2);
	imshow("win6", result);
	//保存处理后的图片
	waitKey(0);
	imwrite("result-2.jpg", result);
	}

效果和python的结果是一致的

小结

1本节主要用kmeans实现了照片底色的更改，主要涉及到了python的一些基础知识和opencv的使用
2.本节给出了python和c++两种方法的使用，也给出了基于界面的编写。大家可以参考一下如何在qt的lable控件中显示图片