关于python:OpenCV实现背景分离可用于更改证件照底色

31次阅读

共计 4232 个字符,预计需要花费 11 分钟才能阅读完成。

include <opencv2/opencv.hpp>

include <iostream>

include <algorithm>

include <time.h>

using namespace cv;
using namespace std;
// 输出参数
struct Inputparama {

int thresh = 30;                               // 背景辨认阈值,该值越小,则辨认非背景区面积越大,需有适合范畴,目前为 5 -60
int transparency = 255;                        // 背景替换色透明度,255 为实,0 为通明
int size = 7;                                  // 非背景区边缘虚化参数,该值越大,则边缘虚化水平越显著
cv::Point p = cv::Point(0, 0);                 // 背景色采样点,可通过人机交互获取,也可用默认 (0,0) 点色彩作为背景色
cv::Scalar color = cv::Scalar(255, 255, 255);  // 背景色

};
cv::Mat BackgroundSeparation(cv::Mat src, Inputparama input);
void Clear_MicroConnected_Areas(cv::Mat src, cv::Mat &dst, double min_area);
// 计算差值均方根
int geiDiff(uchar b,uchar g,uchar r,uchar tb,uchar tg,uchar tr)
{

return  int(sqrt(((b - tb)*(b - tb) + (g - tg)*(g - tg) + (r - tr)*(r - tr))/3));

}
int main()
{

cv::Mat src = imread("111.jpg");
Inputparama input;
input.thresh = 100;
input.transparency = 255;
input.size = 6;
input.color = cv::Scalar(0, 0, 255);
clock_t s, e;
s = clock();
cv::Mat result = BackgroundSeparation(src, input);
e = clock();
double dif = e - s;
cout << "time:" << dif << endl;
imshow("original", src);
imshow("result", result);
imwrite("result1.png", result);
waitKey(0);
return 0;

}
// 背景拆散
cv::Mat BackgroundSeparation(cv::Mat src, Inputparama input)
{

cv::Mat bgra, mask;
// 转化为 BGRA 格局,带透明度,4 通道
cvtColor(src, bgra, COLOR_BGR2BGRA);
mask = cv::Mat::zeros(bgra.size(), CV_8UC1);
int row = src.rows;
int col = src.cols;
// 异样数值修改
input.p.x = max(0, min(col, input.p.x));
input.p.y = max(0, min(row, input.p.y));
input.thresh = max(5, min(200, input.thresh));
input.transparency = max(0, min(255, input.transparency));
input.size = max(0, min(30, input.size));
// 确定背景色
uchar ref_b = src.at<Vec3b>(input.p.y, input.p.x)[0];
uchar ref_g = src.at<Vec3b>(input.p.y, input.p.x)[1];
uchar ref_r = src.at<Vec3b>(input.p.y, input.p.x)[2];
// 计算蒙版区域(掩膜)for (int i = 0; i < row; ++i)
{uchar *m = mask.ptr<uchar>(i);
    uchar *b = src.ptr<uchar>(i);
    for (int j = 0; j < col; ++j)
    {if ((geiDiff(b[3*j],b[3*j+1],b[3*j+2],ref_b,ref_g,ref_r)) >input.thresh)
        {m[j] = 255;
        }
    }
}
cv::Mat tmask = cv::Mat::zeros(row + 50, col + 50, CV_8UC1);
mask.copyTo(tmask(cv::Range(25, 25 + mask.rows), cv::Range(25, 25 + mask.cols)));
// 寻找轮廓,作用是填充轮廓内黑洞
vector<vector<Point>> contour;
vector<Vec4i> hierarchy;
// RETR_TREE 以网状结构提取所有轮廓,CHAIN_APPROX_NONE 获取轮廓的每个像素
findContours(tmask, contour, hierarchy, RETR_EXTERNAL, CHAIN_APPROX_NONE);
drawContours(tmask, contour, -1, Scalar(255), FILLED,16);
// 黑帽运算获取同背景色相似的区域,[商品期货](https://www.gendan5.com/futures/cf.html)辨认后填充
cv::Mat hat;
cv::Mat element = getStructuringElement(MORPH_ELLIPSE, Size(31, 31));
cv::morphologyEx(tmask, hat, MORPH_BLACKHAT, element);
hat.setTo(255, hat > 0);
cv::Mat hatd;
Clear_MicroConnected_Areas(hat, hatd, 450);
tmask = tmask + hatd;
mask = tmask(cv::Range(25, 25 + mask.rows), cv::Range(25, 25 + mask.cols)).clone();
// 掩膜滤波,是为了边缘虚化
cv::blur(mask, mask, Size(2 * input.size+1, 2 * input.size + 1));
// 改色
for (int i = 0; i < row; ++i)
{uchar *r = bgra.ptr<uchar>(i);
    uchar *m = mask.ptr<uchar>(i);
    for (int j = 0; j < col; ++j)
    {
        // 蒙版为 0 的区域就是规范背景区
        if (m[j] == 0)
        {r[4 * j] = uchar(input.color[0]);
            r[4 * j + 1] = uchar(input.color[1]);
            r[4 * j + 2] = uchar(input.color[2]);
            r[4 * j + 3] = uchar(input.transparency);
        }
        // 不为 0 且不为 255 的区域是轮廓区域(边缘区),须要虚化解决
        else if (m[j] != 255)
        {
            // 边缘处按比例上色
            int newb = (r[4 * j] * m[j] * 0.3 + input.color[0] * (255 - m[j])*0.7) / ((255 - m[j])*0.7+ m[j] * 0.3);
            int newg = (r[4 * j+1] * m[j] * 0.3 + input.color[1] * (255 - m[j])*0.7) / ((255 - m[j])*0.7 + m[j] * 0.3);
            int newr = (r[4 * j + 2] * m[j] * 0.3 + input.color[2] * (255 - m[j])*0.7) / ((255 - m[j])*0.7 + m[j] * 0.3);
            int newt = (r[4 * j + 3] * m[j] * 0.3 + input.transparency * (255 - m[j])*0.7) / ((255 - m[j])*0.7 + m[j] * 0.3);
            newb = max(0, min(255, newb));
            newg = max(0, min(255, newg));
            newr = max(0, min(255, newr));
            newt = max(0, min(255, newt));
            r[4 * j] = newb;
            r[4 * j + 1] = newg;
            r[4 * j + 2] = newr;
            r[4 * j + 3] = newt;
        }
    }
}
return bgra;

}
void Clear_MicroConnected_Areas(cv::Mat src, cv::Mat &dst, double min_area)
{

// 备份复制
dst = src.clone();
std::vector<std::vector<cv::Point> > contours;  // 创立轮廓容器
std::vector<cv::Vec4i>     hierarchy;
// 寻找轮廓的函数
// 第四个参数 CV_RETR_EXTERNAL,示意寻找最外围轮廓
// 第五个参数 CV_CHAIN_APPROX_NONE,示意保留物体边界上所有间断的轮廓点到 contours 向量内
cv::findContours(src, contours, hierarchy, cv::RETR_EXTERNAL, cv::CHAIN_APPROX_NONE, cv::Point());
if (!contours.empty() && !hierarchy.empty())
{std::vector<std::vector<cv::Point> >::const_iterator itc = contours.begin();
    // 遍历所有轮廓
    while (itc != contours.end())
    {
        // 定位以后轮廓所在位置
        cv::Rect rect = cv::boundingRect(cv::Mat(*itc));
        // contourArea 函数计算连通区面积
        double area = contourArea(*itc);
        // 若面积小于设置的阈值
        if (area < min_area)
        {
            // 遍历轮廓所在位置所有像素点
            for (int i = rect.y; i < rect.y + rect.height; i++)
            {uchar *output_data = dst.ptr<uchar>(i);
                for (int j = rect.x; j < rect.x + rect.width; j++)
                {
                    // 将连通区的值置 0
                    if (output_data[j] == 255)
                    {output_data[j] = 0;
                    }
                }
            }
        }
        itc++;
    }
}

}

正文完
 0