An Enhanced Adaptive Sobel Edge Detector Based on Improved Genetic Algorithm and Non-Maximum Suppression

This paper presents an enhanced adaptive Sobel edge detector based on improved genetic algorithm (GA) and Non-Maximum Suppression. Three key techniques are presented for increasing positioning precision and noise immunity. Considering the best balance between mean value calculation and differential processing, a new 5*5 operator template is defined to effectively match with the filter template and reduce noise. As the edges gotten by Sobel operator are wide and vivid, we use Non-Maximum Suppression algorithm to maintain the local maximum gradient and eliminate all other gradient values to accomplish the effect of edge thinning. Then we propose an improved GA to achieve threshold adaptation and effectively reduce the probability of this approach troubled into the local optimal threshold. The enhanced adaptive Sobel edge detection algorithm is compared with canny, Sobel and Roberts three traditional edge detection operators and eight direction Sobel operator respectively. We evaluate these five different algorithms by comparing the experimental effect of edge detection, PSNR and time cost. The consequence indicate that the proposed enhanced adaptive Sobel edge detector is better than other compared algorithms in accuracy, denoising ability and time cost.