Removed parallel version for CV_16U label type

modules/imgproc/include/opencv2/imgproc.hpp

@@ -3677,7 +3677,7 @@ the source image. ccltype specifies the connected components labeling algorithm
 Grana (BBDT) and Wu's (SAUF) algorithms are supported, see the cv::ConnectedComponentsAlgorithmsTypes
 for details. Note that SAUF algorithm forces a row major ordering of labels while BBDT does not.
 This function uses parallel version of both Grana and Wu's algorithms if at least one allowed
-parallel framework is enabled.
+parallel framework is enabled and if the rows of the image are at least twice the number returned by getNumberOfCPUs.
 
 @param image the 8-bit single-channel image to be labeled
 @param labels destination labeled image
@@ -3709,7 +3709,7 @@ the source image. ccltype specifies the connected components labeling algorithm
 Grana's (BBDT) and Wu's (SAUF) algorithms are supported, see the cv::ConnectedComponentsAlgorithmsTypes
 for details. Note that SAUF algorithm forces a row major ordering of labels while BBDT does not.
 This function uses parallel version of both Grana and Wu's algorithms (statistics included) if at least one allowed
-parallel framework is enabled.
+parallel framework is enabled and if the rows of the image are at least twice the number returned by getNumberOfCPUs.
 
 @param image the 8-bit single-channel image to be labeled
 @param labels destination labeled image

modules/imgproc/src/connectedcomponents.cpp

@@ -288,7 +288,7 @@ namespace cv{
                 int r = range.start;
                 chunksSizeAndLabels_[r] = range.end;
 
-                LabelT label = LabelT(r  *imgLabels_.cols / 4 + 1);
+                LabelT label = LabelT((r * imgLabels_.cols + 1) / 2 + 1);
 
                 const LabelT firstLabel = label;
                 const int w = img_.cols;
@@ -390,7 +390,7 @@ namespace cv{
                 int r = range.start;
                 chunksSizeAndLabels_[r] = range.end;
 
-                LabelT label = LabelT(r  *imgLabels_.cols / 4 + 1);
+                LabelT label = LabelT((r * imgLabels_.cols + 1) / 2 + 1);
 
                 const LabelT firstLabel = label;
                 const int w = img_.cols;
@@ -606,6 +606,7 @@ namespace cv{
             P[0] = 0;
 
             cv::Range range(0, h);
+            LabelT nLabels = 1;
 
             if (connectivity == 8){
                 //First scan, each thread works with chunk of img.rows/nThreads rows
@@ -623,10 +624,9 @@ namespace cv{
                 //merge labels of different chunks
                 mergeLabels4Connectivity(imgLabels, P, chunksSizeAndLabels);
             }
-            LabelT nLabels = 1;
 
             for (int i = 0; i < h; i = chunksSizeAndLabels[i]){
-                flattenL(P, i  * w / 4 + 1, chunksSizeAndLabels[i + 1], nLabels);
+                flattenL(P, (i * w + 1) / 2 + 1, chunksSizeAndLabels[i + 1], nLabels);
             }
 
             //Array for statistics dataof threads
@@ -650,7 +650,6 @@ namespace cv{
     //Kesheng Wu, et al
     template<typename LabelT, typename PixelT, typename StatsOp = NoOp >
     struct LabelingWu{
-
         LabelT operator()(const cv::Mat& img, cv::Mat& imgLabels, int connectivity, StatsOp& sop){
             CV_Assert(imgLabels.rows == img.rows);
             CV_Assert(imgLabels.cols == img.cols);
@@ -842,7 +841,7 @@ namespace cv{
 
                 chunksSizeAndLabels_[r] = range.end + (range.end % 2);
 
-                LabelT label = LabelT(r  *imgLabels_.cols / 4 + 1);
+                LabelT label = LabelT((r + 1) * (imgLabels_.cols + 1) / 4);
 
                 const LabelT firstLabel = label;
                 const int h = img_.rows, w = img_.cols;
@@ -2562,7 +2561,7 @@ namespace cv{
 
             LabelT nLabels = 1;
             for (int i = 0; i < h; i = chunksSizeAndLabels[i]){
-                flattenL(P, i  *w / 4 + 1, chunksSizeAndLabels[i + 1], nLabels);
+                flattenL(P, (i + 1) * (w + 1) / 4, chunksSizeAndLabels[i + 1], nLabels);
             }
 
             //Array for statistics data
@@ -3932,30 +3931,28 @@ namespace cv{
         int lDepth = L.depth();
         int iDepth = I.depth();
         const char *currentParallelFramework = cv::currentParallelFramework();
+        const int numberOfCPUs = cv::getNumberOfCPUs();
 
         CV_Assert(iDepth == CV_8U || iDepth == CV_8S);
 
+        //Run parallel labeling only if the rows of the image are at least twice the number returned by getNumberOfCPUs
+        const bool is_parallel = currentParallelFramework != NULL && numberOfCPUs > 1 && L.rows / numberOfCPUs >= 2;
+
         if (ccltype == CCL_WU || connectivity == 4){
             // Wu algorithm is used
             using connectedcomponents::LabelingWu;
             using connectedcomponents::LabelingWuParallel;
             //warn if L's depth is not sufficient?
             if (lDepth == CV_8U){
-                if (currentParallelFramework == NULL)
+                //Not supported yet
-                    return (int)LabelingWu<uchar, uchar, StatsOp>()(I, L, connectivity, sop);
-                else
-                    return (int)LabelingWuParallel<uchar, uchar, StatsOp>()(I, L, connectivity, sop);
             }
             else if (lDepth == CV_16U){
-                if (currentParallelFramework == NULL)
                 return (int)LabelingWu<ushort, uchar, StatsOp>()(I, L, connectivity, sop);
-                else
-                    return (int)LabelingWuParallel<ushort, uchar, StatsOp>()(I, L, connectivity, sop);
             }
             else if (lDepth == CV_32S){
                 //note that signed types don't really make sense here and not being able to use unsigned matters for scientific projects
                 //OpenCV: how should we proceed?  .at<T> typechecks in debug mode
-                if (currentParallelFramework == NULL)
+                if (!is_parallel)
                     return (int)LabelingWu<int, uchar, StatsOp>()(I, L, connectivity, sop);
                 else
                     return (int)LabelingWuParallel<int, uchar, StatsOp>()(I, L, connectivity, sop);
@@ -3967,21 +3964,15 @@ namespace cv{
             using connectedcomponents::LabelingGranaParallel;
             //warn if L's depth is not sufficient?
             if (lDepth == CV_8U){
-                if (currentParallelFramework == NULL)
+                //Not supported yet
-                    return (int)LabelingGrana<uchar, uchar, StatsOp>()(I, L, connectivity, sop);
-                else
-                    return (int)LabelingGranaParallel<uchar, uchar, StatsOp>()(I, L, connectivity, sop);
             }
             else if (lDepth == CV_16U){
-                if (currentParallelFramework == NULL)
                 return (int)LabelingGrana<ushort, uchar, StatsOp>()(I, L, connectivity, sop);
-                else
-                    return (int)LabelingGranaParallel<ushort, uchar, StatsOp>()(I, L, connectivity, sop);
             }
             else if (lDepth == CV_32S){
                 //note that signed types don't really make sense here and not being able to use unsigned matters for scientific projects
                 //OpenCV: how should we proceed?  .at<T> typechecks in debug mode
-                if (currentParallelFramework == NULL)
+                if (!is_parallel)
                     return (int)LabelingGrana<int, uchar, StatsOp>()(I, L, connectivity, sop);
                 else
                     return (int)LabelingGranaParallel<int, uchar, StatsOp>()(I, L, connectivity, sop);