Enhanced Automatic Image Parameter setting and Segmentation Method

Abstract

There are a lot of works done to automatically set parameters and segment images based on Pulse Coupled Neural Networks (PCNN). In this study we propose an automatic parameters setting and segmentation method based on Intersecting Cortical Mode (ICM) which enables to overcome the basic limitation of PCNN based methods. We used the ICM as base and developed an enhanced automatic method which can withstand effects of multiple background and illumination during segmentation. Characteristics pixel values of the input image are used to deduce corresponding segmentation parameters. The experiment is done on Aerial Image Segmentation Dataset and Database of Human Segmented Natural Images. Our method outperformed for subjective and objective evaluations, also shown consistent assignment of parameter values. Also the proposed method is able to reduce the segmentation time by half and overcome the limitations of the existing automatic models.