Multitextured Segmentation for Improving Moving Objects Detection and Tracking

Abstract

Image segmentation based on multiple texture features has significant issues in the areas of content based image extraction, image outline recognition, medical image processing, remote sensing, image segmentation through pattern identification and monitoring in crowded public places. Active contour color recognition methods were developed for detecting and tracking object in sequential images. However, the presence of dynamic shadows was a critical issue in foreground segmentation. Therefore, Multi Textured-based Object Segmentation (MTOS) technique is proposed in this study for improving the detection and tracking of moving objects. The proposed technique first locates the objects and boundaries of images with the same label distributed with certain visual characteristics. Next, preprocessing technique is performed using median filtering to reduce the distortion and noise in video frames. Then, texture-based segmentation is carried out using an adaptive threshold-based approach to avoid distortions while detecting moving objects. Detecting moving regions is accomplished by comparing the current video frame from a reference background in a pixel-by-pixel manner with multiple texture features. The effectiveness of moving object image segmentation through texture features is evaluated. The experimental results show that our proposed technique performs better in terms of segmentation accuracy, segmentation time, peak signal to noise ratio and object detection rate.