Authors : S. Anand, R. Shantha Selva Kumari and H. Abirami
Abstract: This study presents a new Edge Detection (ED) in images using simplified Second order Differential Hyperbolic Tangent (SDHT) steerable filter. Smoothing the image with Gaussian filter is the 1st step in conventional ED algorithm. However, shape or smooth parameter of the Gaussian filter influences the overall performance of ED. In particular when the smoothing parameter is >1, Gaussian filter provides excess smoothness and get widened and dislocated edges. This new SDHT filter provides better trade-off between a level of smoothing and edge feature detection than Gaussian filter. This new algorithm not only captures multiple orientations but also well in ED, improved noise tolerance with less computational cost. Mean Squared Error (MSE) and Peak Signal to Noise Ratio (PSNR) parameters are used to measure the quantitative performance of detection and noise tolerance. Experimental results show that the average MSE of 44 different images is reduced in proposed algorithm (0.299975) while compared with Canny (0.30408) and Sobel (0.33152). While filtering the edge information, PSNR is improved in proposed SDHT (57.5) than Gaussian (56.0).
S. Anand, R. Shantha Selva Kumari and H. Abirami, 2012. Edge Detection Using Simplified Second Order Differential Hyperbolic Tangent Steerable Filter. Journal of Engineering and Applied Sciences, 7: 331-338.