E-BiT: Extended Bio-Inspired Texture Descriptor for 2D Texture Analysis and Characterization

This paper presents an extended bio-inspired texture (E-BiT) descriptor for image characterization. The E-BiT combines global ecological concepts of species diversity, evenness, richness, and taxonomic indexes to effectively capture patterns at local levels while maintaining invariance scale, translation, permutation. First, we pre-processed the images by normalizing applying geometric transformations assess properties proposed descriptor. Next, assessed performance on four datasets, including histopathological natural images. Finally, compared it with original BiT other descriptors, such as Haralick, GLCM, LBP. achieved state-of-the-art classification performance, accuracy improvements ranging from 0.12% 20% over descriptors. In addition, demonstrated its generic nature performing well in both histopathologic Future work could examine descriptor’s behavior different spatial scales resolutions optimize property extraction improve performance.