Batik Image Retrieval Using Maximum Run Length LBP and Sine-Cosine Optimizer

This paper investigates the effectiveness of Maximum Run Length (MRL) from Local Binary Pattern (LBP) for content-based Batik image retrieval system. The MRL of state ‘1’ derived from LBP code has been proven to yield a promising result for image classification task. The features generated from MRL of state ‘1’ have lower dimensionality compared to that of the classical LBP feature even though they are from an identical binary string. The additional color features are added for achieving a high image retrieval accuracy. To further increase the accuracy, the Sine-Cosine Algorithm (SCA) and its variants are further applied in the system to obtain the Optimum Similarity Weights (OSW). As documented in Experimental Section, the fusion of MRL and color features along with OSW significantly improves the retrieval accuracy with low feature dimensionality.