Merge pull request #27033 from CodeLinaro:xuezha_3rdPost

quic-xuezha · web-flow · commit 797068853f82 · 2025-03-10T10:24:28.000+03:00
Fix assert failure in Sobel test when enable FastCV #27033 ### Pull Request Readiness Checklist See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request - [x] I agree to contribute to the project under Apache 2 License. - [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV - [ ] The PR is proposed to the proper branch - [ ] There is a reference to the original bug report and related work - [ ] There is accuracy test, performance test and test data in opencv_extra repository, if applicable Patch to opencv_extra has the same branch name. - [ ] The feature is well documented and sample code can be built with the project CMake
diff --git a/3rdparty/fastcv/include/fastcv_hal_imgproc.hpp b/3rdparty/fastcv/include/fastcv_hal_imgproc.hpp
@@ -247,15 +247,15 @@ int fastcv_hal_cvtBGRtoYUVApprox(
 /// @param width Source image width
 /// @param height Source image height
 /// @param cn Number of channels
-/// @param lowThreshold low hresholds value
+/// @param lowThreshold low thresholds value
 /// @param highThreshold high thresholds value
 /// @param ksize Kernel size for Sobel operator.
 /// @param L2gradient Flag, indicating use of L2 or L1 norma.
 ////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
 int fastcv_hal_canny(
-    const uchar* 	src_data,
+    const uchar*    src_data,
     size_t          src_step,
-    uchar* 			dst_data,
+    uchar*          dst_data,
     size_t          dst_step,
     int             width,
     int             height,
diff --git a/3rdparty/fastcv/src/fastcv_hal_imgproc.cpp b/3rdparty/fastcv/src/fastcv_hal_imgproc.cpp
@@ -75,13 +75,13 @@ class FcvSobelLoop_Invoker : public cv::ParallelLoopBody
         // Need additional lines to be border.
         if(range.start > 0)
         {
-            topLines     += halfKernelSize;
-            paddedHeight += halfKernelSize;
+            topLines     += MIN(range.start, halfKernelSize);
+            paddedHeight += MIN(range.start, halfKernelSize);
         }
 
         if(range.end < height)
         {
-            paddedHeight += halfKernelSize;
+            paddedHeight += MIN(height-range.end, halfKernelSize);
         }
 
         cv::Mat srcPadded = src(cv::Rect(0, range.start-topLines, width, paddedHeight));
@@ -305,8 +305,9 @@ int fastcv_hal_sobel(
     else
     {
         int nThreads = cv::getNumThreads();
-        int nStripes = nThreads > 1 ? 3*nThreads : 1;
-
+        // In each stripe, the height should be equal or larger than ksize.
+        // Use 3*nThreads stripes to avoid too many threads.
+        int nStripes = nThreads > 1 ? MIN(height / (int)ksize, 3 * nThreads) : 1;
         cv::parallel_for_(cv::Range(0, height), FcvSobelLoop_Invoker(src, dst, dx, dy, ksize, fcvBorder, 0), nStripes);
     }
 
@@ -1017,10 +1018,10 @@ int fastcv_hal_canny(
 
     if (lowThreshold > highThreshold)
         CV_HAL_RETURN_NOT_IMPLEMENTED("lowThreshold is greater then highThreshold");
-	
-	const double epsilon = 1e-9;
-	
-	if (std::abs(lowThreshold - std::round(lowThreshold)) > epsilon || std::abs(highThreshold - std::round(highThreshold)) > epsilon)
+
+    const double epsilon = 1e-9;
+
+    if (std::abs(lowThreshold - std::round(lowThreshold)) > epsilon || std::abs(highThreshold - std::round(highThreshold)) > epsilon)
         CV_HAL_RETURN_NOT_IMPLEMENTED("threshold with decimal values not supported");
 
     INITIALIZATION_CHECK;
