Optimization of Edge Detection using AVX Intrinsics on Multi-core Architectures

This paper presents the algorithm for augmenting the processing speed of Sobel and Canny edge detection. By reducing the number of arithmetic operations and data loads, the processing time is reduced. Our proposed method is purely based on software approach which does not require any accelerated hardware. In addition, the processing speed is further increased by utilizing the AVX intrinsics and OpenMP. Our proposed Sobel edge detection is on average 28.29 times faster than the Sobel function provided by the OpenCV library. When applied with the Canny edge detection, our algorithm can augment the speed of OpenCV's Canny edge detection by 3.73 percent.