我正在iOS项目中使用OpenCV进行文档边缘检测，并成功检测到文档边缘。
现在，我想旋转图像以及检测到的矩形。我已经提到了 Github project来检测边缘。 为此，我首先旋转图像，并尝试通过再次找到图像的最大矩形来重新检测边缘。不幸的是，它没有给我确切的矩形。
我是否可以建议我一些方法来检测旋转文档的边缘，还是应该将检测到的矩形与图像一起旋转？
+(NSMutableArray *) getLargestSquarePoints: (UIImage *) image : (CGSize) size {
Mat imageMat;
CGColorSpaceRef colorSpace = CGImageGetColorSpace(image.CGImage);
CGFloat cols = image.size.width;
CGFloat rows = image.size.height;
cv::Mat cvMat(rows, cols, CV_8UC4); // 8 bits per component, 4 channels
CGContextRef contextRef = CGBitmapContextCreate(cvMat.data, cols, rows, 8, cvMat.step[0], colorSpace, kCGImageAlphaNoneSkipLast | kCGBitmapByteOrderDefault);
CGContextDrawImage(contextRef, CGRectMake(0, 0, cols, rows), image.CGImage);
CGContextRelease(contextRef);
imageMat = cvMat;
cv::resize(imageMat, imageMat, cvSize(size.width, size.height));
// UIImageToMat(image, imageMat);
std::vector<std::vector<cv::Point> >rectangle;
std::vector<cv::Point> largestRectangle;
getRectangles(imageMat, rectangle);
getlargestRectangle(rectangle, largestRectangle);
if (largestRectangle.size() == 4)
{
// Thanks to: https://stackoverflow.com/questions/20395547/sorting-an-array-of-x-and-y-vertice-points-ios-objective-c/20399468#20399468
NSArray *points = [NSArray array];
points = @[
[NSValue valueWithCGPoint:(CGPoint){(CGFloat)largestRectangle[0].x, (CGFloat)largestRectangle[0].y}],
[NSValue valueWithCGPoint:(CGPoint){(CGFloat)largestRectangle[1].x, (CGFloat)largestRectangle[1].y}],
[NSValue valueWithCGPoint:(CGPoint){(CGFloat)largestRectangle[2].x, (CGFloat)largestRectangle[2].y}],
[NSValue valueWithCGPoint:(CGPoint){(CGFloat)largestRectangle[3].x, (CGFloat)largestRectangle[3].y}] ];
CGPoint min = [points[0] CGPointValue];
CGPoint max = min;
for (NSValue *value in points) {
CGPoint point = [value CGPointValue];
min.x = fminf(point.x, min.x);
min.y = fminf(point.y, min.y);
max.x = fmaxf(point.x, max.x);
max.y = fmaxf(point.y, max.y);
}
CGPoint center = {
0.5f * (min.x + max.x),
0.5f * (min.y + max.y),
};
NSLog(@"center: %@", NSStringFromCGPoint(center));
NSNumber *(^angleFromPoint)(id) = ^(NSValue *value){
CGPoint point = [value CGPointValue];
CGFloat theta = atan2f(point.y - center.y, point.x - center.x);
CGFloat angle = fmodf(M_PI - M_PI_4 + theta, 2 * M_PI);
return @(angle);
};
NSArray *sortedPoints = [points sortedArrayUsingComparator:^NSComparisonResult(id a, id b) {
return [angleFromPoint(a) compare:angleFromPoint(b)];
}];
NSLog(@"sorted points: %@", sortedPoints);
NSMutableArray *squarePoints = [[NSMutableArray alloc] init];
[squarePoints addObject: [sortedPoints objectAtIndex:0]];
[squarePoints addObject: [sortedPoints objectAtIndex:1]];
[squarePoints addObject: [sortedPoints objectAtIndex:2]];
[squarePoints addObject: [sortedPoints objectAtIndex:3]];
imageMat.release();
return squarePoints;
}
else{
imageMat.release();
return nil;
}
}
void getRectangles(cv::Mat& image, std::vector<std::vector<cv::Point>>&rectangles) {
// blur will enhance edge detection
cv::Mat blurred(image);
GaussianBlur(image, blurred, cvSize(11,11), 0);
cv::Mat gray0(blurred.size(), CV_8U), gray;
std::vector<std::vector<cv::Point> > contours;
// find squares in every color plane of the image
for (int c = 0; c < 3; c++)
{
int ch[] = {c, 0};
mixChannels(&blurred, 1, &gray0, 1, ch, 1);
// try several threshold levels
const int threshold_level = 2;
for (int l = 0; l < threshold_level; l++)
{
// Use Canny instead of zero threshold level!
// Canny helps to catch squares with gradient shading
if (l == 0)
{
Canny(gray0, gray, 10, 20, 3); //
// Canny(gray0, gray, 0, 50, 5);
// Dilate helps to remove potential holes between edge segments
dilate(gray, gray, cv::Mat(), cv::Point(-1,-1));
}
else
{
gray = gray0 >= (l+1) * 255 / threshold_level;
}
// Find contours and store them in a list
findContours(gray, contours, CV_RETR_LIST, CV_CHAIN_APPROX_SIMPLE);
// Test contours
std::vector<cv::Point> approx;
for (size_t i = 0; i < contours.size(); i++)
{
// approximate contour with accuracy proportional
// to the contour perimeter
approxPolyDP(cv::Mat(contours[i]), approx, arcLength(cv::Mat(contours[i]), true)*0.02, true);
// Note: absolute value of an area is used because
// area may be positive or negative - in accordance with the
// contour orientation
if (approx.size() == 4 &&
fabs(contourArea(cv::Mat(approx))) > 1000 &&
isContourConvex(cv::Mat(approx)))
{
double maxCosine = 0;
for (int j = 2; j < 5; j++)
{
double cosine = fabs(angle(approx[j%4], approx[j-2], approx[j-1]));
maxCosine = MAX(maxCosine, cosine);
}
if (maxCosine < 0.3)
rectangles.push_back(approx);
}
}
}
}
}
void getlargestRectangle(const std::vector<std::vector<cv::Point> >&rectangles, std::vector<cv::Point>& largestRectangle)
{
if (!rectangles.size())
{
return;
}
double maxArea = 0;
int index = 0;
for (size_t i = 0; i < rectangles.size(); i++)
{
cv::Rect rectangle = boundingRect(cv::Mat(rectangles[i]));
double area = rectangle.width * rectangle.height;
if (maxArea < area)
{
maxArea = area;
index = i;
}
}
largestRectangle = rectangles[index];
}
double angle(cv::Point pt1, cv::Point pt2, cv::Point pt0) {
double dx1 = pt1.x - pt0.x;
double dy1 = pt1.y - pt0.y;
double dx2 = pt2.x - pt0.x;
double dy2 = pt2.y - pt0.y;
return (dx1*dx2 + dy1*dy2)/sqrt((dx1*dx1 + dy1*dy1)*(dx2*dx2 + dy2*dy2) + 1e-10);
}
+(UIImage *) getTransformedImage: (CGFloat) newWidth : (CGFloat) newHeight : (UIImage *) origImage : (CGPoint [4]) corners : (CGSize) size {
cv::Mat imageMat;
CGColorSpaceRef colorSpace = CGImageGetColorSpace(origImage.CGImage);
CGFloat cols = size.width;
CGFloat rows = size.height;
cv::Mat cvMat(rows, cols, CV_8UC4); // 8 bits per component, 4 channels
CGContextRef contextRef = CGBitmapContextCreate(cvMat.data,
// Pointer to backing data
cols,
// Width of bitmap
rows,
// Height of bitmap
8,
// Bits per component
cvMat.step[0],
// Bytes per row
colorSpace,
// Colorspace
kCGImageAlphaNoneSkipLast |
kCGBitmapByteOrderDefault); // Bitmap info flags
CGContextDrawImage(contextRef, CGRectMake(0, 0, cols, rows), origImage.CGImage);
CGContextRelease(contextRef);
imageMat = cvMat;
cv::Mat newImageMat = cv::Mat( cvSize(newWidth,newHeight), CV_8UC4);
cv::Point2f src[4], dst[4];
src[0].x = corners[0].x;
src[0].y = corners[0].y;
src[1].x = corners[1].x;
src[1].y = corners[1].y;
src[2].x = corners[2].x;
src[2].y = corners[2].y;
src[3].x = corners[3].x;
src[3].y = corners[3].y;
dst[0].x = 0;
dst[0].y = -10;
dst[1].x = newWidth - 1;
dst[1].y = -10;
dst[2].x = newWidth - 1;
dst[2].y = newHeight + 1;
dst[3].x = 0;
dst[3].y = newHeight + 1;
dst[0].x = 0;
dst[0].y = 0;
dst[1].x = newWidth - 1;
dst[1].y = 0;
dst[2].x = newWidth - 1;
dst[2].y = newHeight - 1;
dst[3].x = 0;
dst[3].y = newHeight - 1;
cv::warpPerspective(imageMat, newImageMat, cv::getPerspectiveTransform(src, dst), cvSize(newWidth, newHeight));
//Transform to UIImage
NSData *data = [NSData dataWithBytes:newImageMat.data length:newImageMat.elemSize() * newImageMat.total()];
CGColorSpaceRef colorSpace2;
if (newImageMat.elemSize() == 1) {
colorSpace2 = CGColorSpaceCreateDeviceGray();
} else {
colorSpace2 = CGColorSpaceCreateDeviceGray();
// colorSpace2 = CGColorSpaceCreateDeviceRGB();
}
CGDataProviderRef provider = CGDataProviderCreateWithCFData((__bridge CFDataRef)data);
CGFloat width = newImageMat.cols;
CGFloat height = newImageMat.rows;
CGImageRef imageRef = CGImageCreate(width, height, 8, 8 * newImageMat.elemSize(),
newImageMat.step[0],
colorSpace2,
kCGImageAlphaNone | kCGBitmapByteOrderDefault, provider,
NULL, false, kCGRenderingIntentDefault);
UIImage *image = [[UIImage alloc] initWithCGImage:imageRef];
CGImageRelease(imageRef);
CGDataProviderRelease(provider);
CGColorSpaceRelease(colorSpace2);
return image;
}
答案 0 :(得分：0)
如果使用cv2.minAreaRect，它将为轮廓和角度提供最佳的封闭矩形，因此可以向后旋转。