Paper Details: Downloads: 492
Serial Number: P1151021071
Title: Performance Evaluation of Adaptive Statistical Thresholding Based Edge Detection Using GLCM in Wavelet Domain under Noisy Conditions
Authors: K. Padma Vasavi and N. Udaya Kumar and E. V. Krishna Rao and M. Madhavi Latha
Abstract: In this paper, the method of edge detection with Adaptive Statistical Thresholding (AST) using GLCM in wavelet domain is proposed. A multi scale approach using different wavelets is implemented for good localization and good detection of edges. A Gray Level Co-Occurrence Matrix (GLCM) is used to determine the threshold. A statistical approach is made to determine a threshold that is unique for each image considered. . The performance of the Adaptive Statistical Thresholding (AST) detector is compared with the most popular edge detector like the Canny edge detector and also with a new edge detector that uses a morphological filter for edge detection. The comparison is made by making use of both subjective and objective methods of evaluation. Under noise free conditions the AST method is shown to be performing outstandingly well in terms of its Peak Signal to Noise Ratio (PSNR) as the variation in PSNR in this method is better by a range of 5 dB to 12 dB. It is also superior to other methods by means of appearance of the edge maps. The edge detector is also tested for its robustness in presence of noise. The AST edge detector is performing appreciably well even under noisy conditions.
Keywords: Edge detection, Performance Evaluation, Multi scale decomposition, GLCM, Thresholding, standardization.
Journal/Conference: International Journal of Graphics, Vision and Image Processing
Volume: 10
Issue: 3
Submission Date: 5/15/2010 12:00:00 AM
Review Date: 7/5/2010 12:00:00 AM
Publishing Date: 7/26/2010 12:00:00 AM
Article Downloads: 492
Download:

Facebook