using namespace cv;
+//TODO : All the following methods but ComputeContours use the C API of OpenCV while ComputContours
+// uses the C++ API of the library.
+// This should be homogenized and preferably by using the C++ API (which is more recent for all the methods
+
+// The code has to be "cleaned up" too
+
/*!
Constructor
\param theFilename - image to process
/*!
Computes the contours of the image located at imagePath
*/
-bool OCCViewer_FeatureDetector::ComputeContours(){
+bool OCCViewer_FeatureDetector::ComputeContours( int detection_method ){
// Initialising images
Mat src, src_gray;
Mat detected_edges;
-
- // Thresholds for Canny detector
- int lowThreshold = 100;
- int ratio = 3;
- int kernel_size = 3; // 3,5 or 7
- bool isRectSelected = (rect.width > 1);
- if (isRectSelected)
- src = _colorFiltering();
- else
- {
- src = imread( imagePath.c_str() );
- }
-
+ // Read image
+ src = imread( imagePath.c_str() );
if( !src.data )
- { return false; }
-
- // Convert the image to grayscale
- if (src.channels() == 3)
- cvtColor( src, src_gray, CV_BGR2GRAY );
- else if (src.channels() == 1)
- src_gray = src;
+ return false;
+
+ if ( detection_method == CANNY ) // The problem is that with that filter the detector detects double contours
+ {
+ // Thresholds for Canny detector
+ int lowThreshold = 100;
+ int ratio = 3;
+ int kernel_size = 3; // 3,5 or 7
+
+ // Convert the image to grayscale
+ if (src.channels() == 3)
+ cvtColor( src, src_gray, CV_BGR2GRAY );
+ else if (src.channels() == 1)
+ src_gray = src;
- if(isRectSelected)
- {
- _detectAndRetrieveContours( src_gray );
- }
- else
- {
// Reduce noise with a kernel 3x3
blur( src_gray, detected_edges, Size(3,3) );
-
// Canny detector
- Canny( detected_edges, detected_edges, lowThreshold, lowThreshold*ratio, kernel_size, /*L2gradient =*/true ); // The problem is that with that filter the detector detects double contours
-
- // Retrieve contours on the Canny result
- _detectAndRetrieveContours( detected_edges );
+ Canny( detected_edges, detected_edges, lowThreshold, lowThreshold*ratio, kernel_size, /*L2gradient =*/true );
+ }
+ else if ( detection_method == COLORFILTER )
+ {
+ if ( !rect.width > 1 )
+ return false;
+ detected_edges = _colorFiltering();
}
+ else if ( detection_method == RIDGE_DETECTOR ) // Method adapted for engineering drawings (e.g. watershed functionnality could be used here cf.OpenCV documentation and samples)
+ {
+ // TODO
+ return false;
+ }
+ _detectAndRetrieveContours( detected_edges );
return true;
}
+/*!
+ Computes the lines in the image located at imagePath
+*/
+bool OCCViewer_FeatureDetector::ComputeLines(){
+ MESSAGE("OCCViewer_FeatureDetector::ComputeLines()")
+ // Initialising images
+ Mat src, src_gray, detected_edges, dst;
+
+ src=imread(imagePath.c_str(), 0);
+
+ Canny( src, dst, 50, 200, 3 );
+ HoughLinesP( dst, lines, 1, CV_PI/180, 80, 30, 10 );
+ return true;
+
+}
+
/*!
Stores a region of interest given by user in rect
\param theRect - Region Of Interest of the image located at imagePath
void OCCViewer_FeatureDetector::_detectAndRetrieveContours( Mat src )
{
src = src > 1;
- int method ;//= CV_CHAIN_APPROX_SIMPLE;
-// if ( rect.width > 1 )
- method = CV_CHAIN_APPROX_NONE;//CV_CHAIN_APPROX_TC89_KCOS;//CV_CHAIN_APPROX_TC89_L1;
+ int method = CV_CHAIN_APPROX_NONE;
findContours( src, contours, hierarchy,CV_RETR_CCOMP, method);
- // Other possible approximations CV_CHAIN_APPROX_TC89_L1, CV_CHAIN_APPROX_SIMPLE cf. OpenCV documentation
+ // Other possible approximations CV_CHAIN_APPROX_TC89_KCOS, CV_CHAIN_APPROX_TC89_L1, CV_CHAIN_APPROX_SIMPLE cf. OpenCV documentation
// for precise information
}
Mat OCCViewer_FeatureDetector::_colorFiltering()
{
IplImage* find_image = cvLoadImage(imagePath.c_str(),CV_LOAD_IMAGE_COLOR);
+ // Reduce noise with a kernel 3x3
+ cvSmooth( find_image, find_image, CV_GAUSSIAN, 3, 3 );
+
if ( !rect.width > 1 )
return Mat(find_image);
IplImage* test_hsv = cvCreateImage(cvGetSize(test_image),8,3);
IplImage* test_hue = cvCreateImage(cvGetSize(test_image),8,1);
-
CvHistogram* hist;
cvCvtColor(test_image, test_hsv, CV_BGR2HSV);
//calculate hue` histogram
cvCalcHist(&test_hue, hist, 0 ,0);
+// // TEST print of the histogram for debugging
+// IplImage* hist_image = cvCreateImage(cvSize(320,300),8,3);
+//
+// //draw hist on hist_test image.
+// cvZero(hist_image);
+// float max_value = 0;
+// cvGetMinMaxHistValue(hist, 0 , &max_value, 0, 0);
+// int bin_w = hist_image->width/size_hist;
+// for(int i = 0; i < size_hist; i++ )
+// {
+// //prevent overflow
+// int val = cvRound( cvGetReal1D(hist->bins,i)*hist_image->
+// height/max_value);
+// CvScalar color = CV_RGB(200,0,0);
+// //hsv2rgb(i*180.f/size_hist);
+// cvRectangle( hist_image, cvPoint(i*bin_w,hist_image->height),
+// cvPoint((i+1)*bin_w,hist_image->height - val),
+// color, -1, 8, 0 );
+// }
+//
+//
+// cvNamedWindow("hist", 1); cvShowImage("hist",hist_image);
+
+
//calculate back projection of hue plane of input image
IplImage* backproject = cvCreateImage(cvGetSize(find_image), 8, 1);
IplImage* binary_backproject = cvCreateImage(cvGetSize(find_image), 8, 1);
cvCalcBackProject(&find_hue, backproject, hist);
// Threshold in order to obtain binary image
- cvThreshold(backproject, binary_backproject, 1, 255, CV_THRESH_BINARY); // NOTE it would be good to think about the best threshold to use (1 here)
+ cvThreshold(backproject, binary_backproject, 1, 255, CV_THRESH_BINARY); // NOTE it would be good to think about the best threshold to use (it's 1 for now)
cvReleaseImage(&test_image);
cvReleaseImage(&test_hsv);
cvReleaseImage(&test_hue);
