Asim ur Rehman Khan
National University of Computer & Emerging Sciences, Pakistan
Title: Edge detection of moving objects in the highly corrupted image sequence
Biography
Biography: Asim ur Rehman Khan
Abstract
The detection of important features of a moving object is a challenging task especially when these images are corrupted with heavy noise. This research proposes two statistical base techniques. The first technique performs three-way nested design using the analysis of variance (ANOVA). The three-way nested design corresponds to three-layers. The top layer is based on the temporal analysis where the model compares two consecutive image frames and identifies regions having sufficient temporal interframe changes. The next two layers perform statistical approach to see if there are sufficient intraframe variations. A large amount of intraframe variations are accounted for important features that may have edges to track across multiple image frames. In case of affirmative results in all the three layers, a second method based on the contrast function (CF) is used to identify edges in four possible directions. These four directions are horizontal, vertical, and two diagonal directions. The presence or absence of an effect is confirmed by testing a hypothesis. The test uses F-test, and Tukey’s T-test. The results are quite good for image frames that are previously corrupted with heavy Gaussain noise.
