Sobel-fuzzy technique to enhance the detection of edges in grayscale images using auto-thresholding

Abstract

Images have always been very important in human life because humans are very much adapted in understanding images. Feature points or pixels play very important role in image analysis. These feature points include edge pixels. Edges on the image are strong intensity variations which show the difference between an object and the background. Edge detection is one of the most important operations in image analysis as it helps to reduce the amount of data by filtering out the less relevant information and if edge can be identified, basic properties of object such as area, perimeter, shape, etc can be measured. In this paper, a Sobel-Fuzzy technique using auto-thresholding is proposed by fuzzifying the results of first derivatives of Sobel in x, y and xy directions. The technique automatically finds the six threshold values using local thresholding. Comparative study has been done on the basis of visual perception and edgel counts. The experimental results show the proposed Sobel-Fuzzy approach is more efficient in comparison to Roberts, Prewitt, Sobel, and LoG and produces better results.