注:此教程是对贾志刚老师的opencv课程学习的一个记录,在此表示对贾老师的感谢.
需求: 扫描仪扫描到的法律文件,需要切边,去掉边缘空白
例如下图,左边是要处理的图像,右边是处理过后的图像.
解决思路: 边缘检测----> 轮廓发现或者直线检测最大外接矩形---->截图最大外接矩阵所在的区域
#include <opencv2/opencv.hpp>
#include <iostream>
#include <math.h>
using namespace cv;
using namespace std;
Mat src, gray_src, dst;
int threshold_value = 100;
int max_level = 255;
const char *output_win = "Contours Result";
const char *roi_win = "Final Result";
void FindROI(int, void *);
int main(int argc, char **argv) {
src = imread("/home/fuhong/code/cpp/opencv_learning/src/小案例/imgs/case1.png");
if (src.empty()) {
printf("could not load image...\n");
return -1;
}
namedWindow("input image", CV_WINDOW_AUTOSIZE);
imshow("input image", src);
namedWindow(output_win, CV_WINDOW_AUTOSIZE);
// namedWindow(roi_win, CV_WINDOW_AUTOSIZE);
//createTrackbar("Threshold:", output_win, &threshold_value, max_level, FindROI);
FindROI(0, 0);
waitKey(0);
return 0;
}
void FindROI(int, void *) {
cvtColor(src, gray_src, COLOR_BGR2GRAY); //先将图像转化为灰度图
Mat canny_output;
Canny(gray_src, canny_output, threshold_value, threshold_value * 2, 3, false); //canny边缘检测
vector<vector<Point>> contours;
vector<Vec4i> hireachy;
findContours(canny_output, contours, hireachy, RETR_TREE, CHAIN_APPROX_SIMPLE, Point(0, 0)); //查找轮廓
/*
* // 在二值图像上发现轮廓使用
cv::findContours(
InputOutputArray binImg, // 输入图像,非0的像素被看成1,0的像素值保持不变,8-bit
OutputArrayOfArrays contours,// 全部发现的轮廓对象
OutputArray, hierachy// 图该的拓扑结构,可选,该轮廓发现算法正是基于图像拓扑结构实现。
int mode, // 轮廓返回的模式
int method,// 发现方法
Point offset=Point()// 轮廓像素的位移,默认(0, 0)没有位移
)
*/
int minw = src.cols * 0.75;
int minh = src.rows * 0.75;
RNG rng(12345);
Mat drawImage = Mat::zeros(src.size(), CV_8UC3);
Rect bbox;
for (size_t t = 0; t < contours.size(); t++) {
//找到轮廓的最小外接斜矩形 参考:https://blog.csdn.net/u013925378/article/details/84563011
RotatedRect minRect = minAreaRect(contours[t]);
float degree = abs(minRect.angle);
// 如果矩阵轮廓大于0.75倍的图像长度和宽度,并且不是图像的最外层轮廓,则是要寻找的轮廓
if (minRect.size.width > minw && minRect.size.height > minh && minRect.size.width < (src.cols - 5)) {
printf("current angle : %f\n", degree);
Point2f pts[4];
minRect.points(pts);
bbox = minRect.boundingRect();
Scalar color = Scalar(rng.uniform(0, 255), rng.uniform(0, 255), rng.uniform(0, 255));
for (int i = 0; i < 4; i++) {
line(drawImage, pts[i], pts[(i + 1) % 4], color, 2, 8, 0);
}
}
}
imshow(output_win, drawImage);
if (bbox.width > 0 && bbox.height > 0) {
Mat roiImg = src(bbox);
imshow(roi_win, roiImg);
}
return;
}
效果:
进阶需求:有时候图像不是正的,有一定的倾斜角度,这种情况下,将图像旋正,并且去除白边.
解决思路: 边缘检测----> 轮廓发现或者直线检测最大外接矩形---->计算需旋转的角度----->将图像旋正----->边缘检测----> 轮廓发现或者直线检测最大外接矩形---->截取最大外接矩阵所在的区域
代码实现:
#include <opencv2/opencv.hpp>
#include <iostream>
#include <math.h>
using namespace cv;
using namespace std;
Mat src, gray_src, dst;
int threshold_value = 100;
int max_level = 255;
const char* output_win = "Contours Result";
const char* roi_win = "Final Result";
void FindROI(int, void*);
void Check_Skew(int, void*);
int main(int argc, char** argv) {
src = imread("D:/gloomyfish/case1r.png");
if (src.empty()) {
printf("could not load image...\n");
return -1;
}
namedWindow("input image", CV_WINDOW_AUTOSIZE);
imshow("input image", src);
namedWindow(output_win, CV_WINDOW_AUTOSIZE);
Check_Skew(0, 0);
// namedWindow(roi_win, CV_WINDOW_AUTOSIZE);
//createTrackbar("Threshold:", output_win, &threshold_value, max_level, FindROI);
// FindROI(0, 0);
waitKey(0);
return 0;
}
void Check_Skew(int, void*) {
Mat canny_output;
cvtColor(src, gray_src, COLOR_BGR2GRAY);
Canny(gray_src, canny_output, threshold_value, threshold_value * 2, 3, false);
vector<vector<Point>> contours;
vector<Vec4i> hireachy;
findContours(canny_output, contours, hireachy, RETR_TREE, CHAIN_APPROX_SIMPLE, Point(0, 0));
Mat drawImg = Mat::zeros(src.size(), CV_8UC3);
float maxw = 0;
float maxh = 0;
double degree = 0;
for (size_t t = 0; t < contours.size(); t++) {
RotatedRect minRect = minAreaRect(contours[t]);
degree = abs(minRect.angle);
if (degree > 0) {
maxw = max(maxw, minRect.size.width);
maxh = max(maxh, minRect.size.height);
}
}
RNG rng(12345);
for (size_t t = 0; t < contours.size(); t++) {
RotatedRect minRect = minAreaRect(contours[t]);
if (maxw == minRect.size.width && maxh == minRect.size.height) {
degree = minRect.angle;
Point2f pts[4];
minRect.points(pts);
Scalar color = Scalar(rng.uniform(0, 255), rng.uniform(0, 255), rng.uniform(0, 255));
for (int i = 0; i < 4; i++) {
line(drawImg, pts[i], pts[(i + 1) % 4], color, 2, 8, 0);
}
}
}
printf("max contours width : %f\n", maxw);
printf("max contours height : %f\n", maxh);
printf("max contours angle : %f\n", degree);
imshow(output_win, drawImg);
Point2f center(src.cols / 2, src.rows / 2);
Mat rotm = getRotationMatrix2D(center, degree, 1.0);
Mat dst;
warpAffine(src, dst, rotm, src.size(), INTER_LINEAR, 0, Scalar(255, 255, 255));
imshow("Correct Image", dst);
}
void FindROI(int, void*) {
cvtColor(src, gray_src, COLOR_BGR2GRAY);
Mat canny_output;
Canny(gray_src, canny_output, threshold_value, threshold_value * 2, 3, false);
vector<vector<Point>> contours;
vector<Vec4i> hireachy;
findContours(canny_output, contours, hireachy, RETR_TREE, CHAIN_APPROX_SIMPLE, Point(0, 0));
int minw = src.cols*0.75;
int minh = src.rows*0.75;
RNG rng(12345);
Mat drawImage = Mat::zeros(src.size(), CV_8UC3);
Rect bbox;
for (size_t t = 0; t < contours.size(); t++) {
RotatedRect minRect = minAreaRect(contours[t]);
float degree = abs(minRect.angle);
if (minRect.size.width > minw && minRect.size.height > minh && minRect.size.width < (src.cols-5)) {
printf("current angle : %f\n", degree);
Point2f pts[4];
minRect.points(pts);
bbox = minRect.boundingRect();
Scalar color = Scalar(rng.uniform(0, 255), rng.uniform(0, 255), rng.uniform(0, 255));
for (int i = 0; i < 4; i++) {
line(drawImage, pts[i], pts[(i + 1)%4], color, 2, 8, 0);
}
}
}
imshow(output_win, drawImage);
if (bbox.width > 0 && bbox.height > 0) {
Mat roiImg = src(bbox);
imshow(roi_win, roiImg);
}
return;
}