Using Sliding Window Technique to Explore the Variations of Image Pixels for Edge Detection

Abstract

This paper proposes an edge detection method using sliding window technique. Compared to traditional methods, the major difference is that the size of sliding window is floating. The window size will be modified according to the variance of all pixel values and mean value of the sum of all differences between each pixel value and the minimum pixel value in the sliding window. This dynamic windowing concept will be applied to detect edges from four different directions, hence many windows with different size will be generated. All pixel values in the same window are similar because of their low variations. After all pixels of the image being included into a certain window, four directional edge maps are generated. The proposed method can detect much thinner edges than Sobel method. Compared with Kang and Wang’s method, the proposed method can extract all the major edges and is also timesaving.