Journal of Applied Research in Electrical Engineering

Journal of Applied Research in Electrical Engineering

Texture Classification Using Deep Learning and Enhanced Local Binary Pattern

Document Type : Research article

Authors
1 Department of Electrical Engineering, Ahar Branch, Islamic Azad University, Ahar, Iran
2 Department of Electrical Engineering, Urmia Branch, Islamic Azad University, Urmia, Iran
3 Department of Electrical Engineering, Khoy Branch, Islamic Azad University, Khoy, Iran
Abstract
Image texture classification is crucial in machine vision and image processing. The primary step in this classification process involves extracting features from the image. Numerous techniques have been developed for feature extraction from textured images; however, local binary patterns (LBP), in both their original and enhanced forms, stand out due to their simplicity in implementation and their ability to extract effective features that yield high classification accuracy. Experts widely recognize that deep neural networks excel at classifying images and extracting detailed features. Building on the strengths of these methods, this study introduces a new model that merges deep learning with enhanced local binary patterns. This model has been tested on the Outex and CUReT datasets. The experimental results indicate that combining deep learning with an improved local binary pattern method significantly enhances the accuracy of texture image classification. According to these results, the classification accuracy of this new model surpasses that of previous methods.

Highlights

  • The paper presents a novel model combining enhanced Local Binary Pattern (LBP) with deep neural networks to improve texture image classification accuracy.
  • The proposed model demonstrated superior performance over previous methods, achieving 99% accuracy in texture classification across multiple datasets (Outex and CUReT).
  • The model showed excellent robustness to challenges such as lighting variations, rotation, and scale changes, outperforming other traditional texture classification methods.
  • By integrating the LBP method with deep learning architectures like convolutional neural networks (CNNs), the model effectively extracts and combines texture features, leading to more reliable classification.
  • The proposed method was compared against other texture classification methods and showed consistent improvements in accuracy, with a performance boost ranging from 1% to 10% over previous models.

Keywords
Subjects

 
Volume 4, Issue 1
Winter and Spring 2025
January 2025
Pages 94-105

  • Receive Date 09 July 2024
  • Revise Date 20 April 2025
  • Accept Date 06 August 2025