Interactive Image Segmentation using Edge Point Techniques (EPT) for Background Subtraction and Object Tracking

摘要

This paper focus on developing interactive image segmentation using Edge Point Technique(EPT). There are a lot of Image segmentation techniques are used in image processing such as adaptive constraint propagation, mean shifting techniques, graph based Segmentation, hybrid segmentation etc. The proposed research uses Edge Point Technique (EPT) for Background and Foreground Separation. EPT makes decisions based on particular pixel information and are effective when the moderation levels of the objects fall squarely outside the range of levels in the background. EPT generates pairwise constraints and performs seed propagation. Pairwise constraints in EPT propagate characteristics of the user’s interactive information through the whole image and effectively preserve global discriminative data coherence, thus avoiding bias caused by the limited interactive information. Seed propagation in EPT significantly reduces the computational complexity in interactive image segmentation by decomposing the learning procedure of an image into blocks. The method first extract features from superpixels obtained by existing threshold based segmentation in an image and Pairwise constraints are generated from the user’s interactive information. Next, EPT performs seed propagation on both features and pairwise constraints to learn the global structure in an image. Experimental results demonstrate that the proposed EPT successfully segments foreground objects from the background and remarkably acceptable computational costs.