A Quantitative and Comparative Analysis on Variations of a Method for Segmentation of Similar Images

Segmenting images is one of the most famous image processing and computer vision's challenge. In this work we present an approach to a method of segmentation of similar images based on hierarchical segmentation and texture information. An empirical study was conducted, it was made of three experiments. First experiment evaluated the P param-eter's variation of the LBP texture extractor and the use of external edges of analyzed region; the second one evaluated the variation of eight gradients, being 7 colored gradients; and the last one compares results between Kullback-Leibler measure and Euclidean distance. The obtained results show the best configurations.