Publication

Abstract : Non-uniformly illuminated images are a class of images that, from a subjective perspective, are difficult to analyze. The excess noise and the lack of properly defined boundaries all contribute to making these images a difficult dataset for any form of analysis or segmentation. This calls for proper feature extraction and specific enhancement to make these images ready for efficient information gathering. This paper aims to visualize the features that can be enhanced using image enhancement techniques to identify the target animal in a non-uniformly illuminated and occluded image, thereby enhancing the recognition power of the proposed system. This paper uses a method to approximately detect and locate the position of the animal in an image. Segmentation Using Region Adjacency Graphs, Interactive Foreground Extraction using GrabCut Algorithm and DeepLab model for semantic image segmentation have also been analyzed.