check if threadIdx.x > 0 && threadIdx.y > 0

src/tv_denoising.cu

@@ -21,7 +21,7 @@ __global__ void tvDenoisingKernel(float* image, int width, int height, float lam
      // Perform TV denoising iteratively
      for (int iteration = 0; iteration < maxIterations; ++iteration)
      {
-         // Calculate the gradients using central differences
+        // Calculate the gradients using central differences
         gradientX[threadIdx.x][threadIdx.y] = image[index + 1] - image[index - 1];
         gradientY[threadIdx.x][threadIdx.y] = image[index + width] - image[index - width];
 
@@ -29,10 +29,12 @@ __global__ void tvDenoisingKernel(float* image, int width, int height, float lam
         __syncthreads();
 
         // Apply TV denoising update rule
-        updatedImage[threadIdx.x][threadIdx.y] = image[index] + lambda * (
-            gradientX[threadIdx.x][threadIdx.y] - gradientX[threadIdx.x - 1][threadIdx.y] +
-            gradientY[threadIdx.x][threadIdx.y] - gradientY[threadIdx.x][threadIdx.y - 1]
-        );
+        if (threadIdx.x > 0 && threadIdx.y > 0) {
+            updatedImage[threadIdx.x][threadIdx.y] = image[index] + lambda * (
+                gradientX[threadIdx.x][threadIdx.y] - gradientX[threadIdx.x - 1][threadIdx.y] +
+                gradientY[threadIdx.x][threadIdx.y] - gradientY[threadIdx.x][threadIdx.y - 1]
+            );
+        }
 
         // Update the global image array with the updated pixel values
         image[index] = updatedImage[threadIdx.x][threadIdx.y];
@@ -65,6 +67,12 @@ extern "C" void TVDenoising(cv::Mat& image, float lambda, int maxIterations)
 
     // Invoke the TV denoising kernel
     tvDenoisingKernel<<<gridSize, blockSize>>>(d_image, width, height, lambda, maxIterations);
+    // Check for errors during kernel launch
+    cudaError_t err = cudaGetLastError();
+    if (err != cudaSuccess)
+    {
+        printf("Error: %s\n", cudaGetErrorString(err));
+    }
 
     // Copy the denoised image data back from device to host
     cudaMemcpy(floatImage.ptr<float>(0), d_image, width * height * sizeof(float), cudaMemcpyDeviceToHost);