A Modified Segmentation Approach for Real World Images Based on Edge Density Associated with Image Contrast Stretching

Abstract

Segmentation of real world images considered as one of the most challenging tasks in the computer vision field due to several issues that associated with this kind of images such as high interference between object foreground and background, complicated objects and the pixels intensities of the object and background are almost similar in some cases. This research has introduced a modified adaptive segmentation process with image contrast stretching namely Gamma Stretching to improve the segmentation problem. The iterative segmentation process based on the proposed criteria has given the flexibility to the segmentation process in finding the suitable region of interest. As well as, the using of Gamma stretching will help in separating the pixels of the objects and background through making the dark intensity pixels darker and the light intensity pixels lighter. The first 20 classes of Caltech 101 dataset have been utilized to demonstrate the performance of the proposed segmentation approach. Also, the Saliency Cut method has been adopted as a benchmark segmentation method. In summary, the proposed method improved some of the segmentation problems and outperforms the current segmentation method namely Saliency Cut method with segmentation accuracy 77.368%, as well as it can be used as a very useful step in improving the performance of visual object categorization system because the region of interest is mostly available.