Texture Analysis on Digital Microscopic Leather Images For Species Identification

Abstract

This paper describes the relevance of texture analysis on leather images. The aim is to improve the prediction accuracy by quantifying the morphological and statistical behavior of the leather images. Hence, the present work proposed to combine the multi-resolution discrete wavelet transform (DWT) and local binary pattern (LBP) texture operators. The hybrid texture features (DWT + LBP) offer better species-specific feature discrimination. This work adopts a multi-layer perceptron (MLP) model to evaluate the discriminatory behavior of the texture features. The proposed work extract, analyze and learn the species' distinct texture features of the novel digital microscopic leather image data. The experimental results noted a significant improvement in species prediction with 99.58% accuracy. Therefore, texture analysis elevates the ability to interpret the leather images per species. It is thus a necessary key to learn the permissible leather species' behavior so as to prevent the trade of non-permissible leather and its products.