frotms / line_detector

line segment detector(lsd) &. edge drawing line detector(edl) &. hough line detector(standard &. probabilistic) for detection

Three mainstream line detectors: lsd-lines, ed-lines and hough-lines

• segment line detector (lsd)
• edge drawing line detector (edlines)
• hough line detector (standard and probabilistic)

All original dependencies have been removed. Code could be run independently:

• line segment detector with a scale in vertical and horizontal direction in boundingbox, respectively
• edge drawing line detector with a scale in vertical and horizontal direction in boundingbox, respectively
• hough line detector(standard and probabilistic) with a scale in vertical and horizontal direction in boundingbox, respectively

EDLines

EDLines Simple Interface with Scale in Boundingbox

@param src         				image,single channel.

@param w           				width of image.

@param h           				height of image.

@param scaleX      				downscale factor in X-axis.

@param scaleY      				downscale factor in Y-axis.

@param bbox        				boundingbox to detect.

@param lines      				result.

@return            				0:ok; 1:error
int EdgeDrawingLineDetector(unsigned char *src, int w, int h,float scaleX, scaleY, boundingbox_t bbox, std::vector<line_float_t> &lines);

LSD

LSD Simple Interface with Scale in Boundingbox

@param src         				image,single channel.

@param w           				width of image.

@param h           				height of image.

@param scaleX      				downscale factor in X-axis.

@param scaleY      				downscale factor in Y-axis.

@param bbox       			 	boundingbox to detect.

@param lines       				result.

@return            				0:ok; 1:error
int LsdLineDetector(unsigned char *src, int w, int h, float scaleX, float scaleY, boundingbox_t bbox, std::vector<line_float_t> &lines);

Houghline

Houghline Simple Interface with Scale in Boundingbox

@param src         				image,single channel.

@param w           				width of image.

@param h           				height of image.

@param scaleX      			 	downscale factor in X-axis.

@param scaleY      			 	downscale factor in Y-axis.

@param canny_low_thresh      	lower threshold for the hysteresis procedure in canny operator.

@param canny_high_thresh      	higher threshold for the hysteresis procedure in canny operator.

@param hough_rho      			distance resolution of the accumulator in pixels.

@param hough_theta      		angle resolution of the accumulator in radians.

@param min_theta_linelength     standard: for standard and multi-scale hough transform, minimum angle to check for lines; propabilistic: minimum line length. Line segments shorter than that are rejected.

@param max_theta_gap      		standard: for standard and multi-scale hough transform, maximum angle to check for lines; propabilistic: maximum allowed gap between points on the same line to link them.

@param hough_thresh      		accumulator threshold parameter. only those lines are returned that get enough votes ( >threshold ).

@param _type      				hough line method: HOUGH_LINE_STANDARD or HOUGH_LINE_PROBABILISTIC

@param bbox        				boundingbox to detect.

@param lines       				result.

@return            				0:ok; 1:error
int HoughLineDetector(unsigned char *src, int w, int h, float scaleX, float scaleY, float CannyLowThresh, float CannyHighThresh, float HoughRho, float HoughTheta, float MinThetaLinelength, float MaxThetaGap, int HoughThresh, HOUGH_LINE_TYPE_CODE _type, boundingbox_t bbox, std::vector<line_float_t> &lines);

Reference

• LSD: A Line Segment Detector
• LBD_Descriptor
• OpenCV

line_detector's Issues

memory error

When I use the following image as input, the algorithm meets with a memory error in function
int EDLineDetector::EDline(image_int8u_p image, LineChains &lines, bool smoothed)
It seems to be a memory out of range error.
@frotms Could you please have a try and resolve the problem?

the code can't caculate the descriptor？

The wrong line segment

This is my code：

`void main()
{
	Mat image;
	image = imread("1.png");
	Mat gray_image;
	cvtColor(image, gray_image, CV_BGR2GRAY);
	int W = image.rows;
	int H = image.cols;
	int image_size = W* H;
	unsigned char* input = new unsigned char[image_size];
	memcpy(input, gray_image.data, image_size);
	vector<line_float_t> Lines;
	boundingbox_t Bbox = { 0,0,W,H };
	float scalex =0.5;
	float scaley =0.5;
	int Flag = 0;
	Flag = LsdLineDetector(input, W, H, scalex, scaley, Bbox, Lines);
	cout << Flag << endl;
	for (int i = 0; i < Lines.size(); i++)
	{
		line(image, Point(Lines[i].startx, Lines[i].starty), Point(Lines[i].endx, Lines[i].endy), Scalar(0, 0, 255), 2);
	}
	imshow("image1", image);
	waitKey(0);
	system("pause");
}`

OpenCV version

Dear sir,

I am trying to use opencv c++ to implement paper title "Edge Drawing – A Combined Real-Time Edge/Edge Segment Detector" but I still can not make it done for part "anchor smart routing algorithm". Could you please help me by sharing the opencv version of this paper?

Thank you so much.

I don't know what the problem is

input image:

my code:
int w = img.cols;
int h = img.rows;
boundingbox_t bbox = { 0,0,w,h };
float scaleX = 0.8;
float scaleY = 0.8;
std::vector<line_float_t> lines;

cv::Point2f pt1, pt2;
cv::Mat result = cv::Mat::zeros(h, w, CV_8UC3);

unsigned char * pImg = img.ptr<uchar>(0);

int ret = LsdLineDetector(pImg, w, h, scaleX, scaleY, bbox, lines);
if (ret < 1)
{		
	for (int i = 0; i < lines.size(); i++)
	{
		line_float_t line = lines[i];
		pt1.x = line.startx;
		pt1.y = line.starty;
		pt2.x = line.endx;
		pt2.y = line.endy;
		int width = 2;
		cv::Scalar color(rand() & 255, rand() & 255, rand() & 255);
		cv::line(result, pt1, pt2, color, width, CV_AA);
		
	}
}
imshow("img", result);
waitKey();

result:

question:
I am sure to load a gray image, but the result is not good, what is the problem? parameter?
it is seemed that filter image by gauss can make a good result.

DBL_EPSILON and FLT_MIN are not defined

as stated in title

Project license?

Hello, I'm interested in using this implementation for a project I'm working on, and I don't want to use your code without permission though, so I was wondering if it'd be possible for you to add a license to this repo. Thanks!