Edge detection for illumination varying images using wavelet similarity

This study presents a new method of edge detection (ED) in illumination varying images. The undesirable dependence because of uneven illumination is crucial in image ED. The proposed algorithm avoids this problem by incorporating a new weighted least-square regularisation with the wavelet filter directional responses. In order to obtain the edge information, the similarity between image neighbourhood pixels and wavelet filter is measured. This new method accomplishes the benefits of wavelets, which includes multi-scale, and wider design choice in the filter length. Figure of merit (F) and structural similarity index qualitatively compares its performances. The subjective and objective results describes that this algorithm performs better than other methods.