A new edge detector based on fractional integration

Mohammed Mekideche, Youcef Ferdi

Abstract: In computer vision and image processing, the Canny edge detector algorithm is the most widely implemented from performance point of view. In this paper, attempting to reduce the computational time of this algorithm on skipping the smoothing step, a fractional integral mask (FIM) is introduced and investigated. It has been shown that the smoothing operation can be omitted when simply convolving the image with our new FIM instead of using integer gradient masks. The efficiency of our FIM is interpreted in term of the running time, robustness to noise, and the potentiality of detecting week edges. The results of simulation show how the quality of edge detection can be enhanced when the fractional factor could be fine tuned. The FIM is a prominent tool which can take part in multispectral images segmentation in the field of satellite imaging.