A Comparative Study on Interactive Segmentation Algorithms for Segmentation of Animal Images

Abstract

Throughout this article, we are researching two distinct algorithms to separate animals from animal pictures. Since animals appear in a very complex background and often surrounded by greenery, segmenting an animal from its background is a very challenging task. Animal segmentation further helps the problem of animal identification and classification. Here the animals are segmented using graph cut and similarity region merging techniques. To determine the efficiency of our process, an analysis is carried out on our own dataset of 50 animal types, comprising 5000 images. The various performance measures such as Information Variation, Global Consistency Error, Probabilistic Rand Index, and Boundary Displacement Error are used for the purpose of assessment.