Enhancing Tomato Leaf Disease Detection with Contrast-Limited Adaptive Histogram Equalization and Image Blending - Improved Machine Learning Accuracy

Aug 3, 2026

This study examines the application of Contrast Limited Adaptive Histogram Equalization (CLAHE), an advanced version of Adaptive Histogram Equalization (AHE), and image blending as a preprocessing technique for tomato leaf images to improve machine learning accuracy in agricultural disease detection. CLAHE is used to normalize image contrast and blending to strengthen feature visibility for downstream analysis. Experimental results show accuracy gains up to 4.02% over baseline models using unprocessed images and improved symptom highlighting versus traditional enhancement methods. The findings support timely disease identification to enable better crop management decisions and production efficiency.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages18

Reading Time45 min

AccessFree to download

Introduction

Background and challenges

Motivation from machine learning and preprocessing quality

Proposed approach and research aims

Related Work

Enhancing tomato leaf disease detection with contrast-limited adaptive histogram equalization and image blending for improved machine learning accuracy Nguyen Khanh Nhan1, Duong Huu Thanh1* 1Ho Chi Minh City Open University, Ho Chi Minh City, Vietnam *Corresponding author: [thanh.dh@ou.edu.vn](thanh.dh@ou.edu.vn) | ARTICLE INFO | ABSTRACT | | --- | --- | | DOI:10 .46223/HCMCOUJS. | This study examines the application of Contrast Limited | | tech.en.16.1.3808.2026 | Adaptive Histogram Equalization (CLAHE), an advanced version of Adaptive Histogram Equalization (AHE), and image blending as a preprocessing technique for tomato leaf images to improve the accuracy of machine learning models in agricultural applications, particularly in the context of disease detection. We implemented CLAHE to normalize the contrast in tomato images and the image blending technique, thereby enhancing the visibility of key features critical for accurate analysis. The experimental results demonstrate a significant increase in the accuracy of the machine learning algorithms, with improvements | | Received: October 20th, 2024 | of up to 4.02% compared to baseline models using standard | | Revised: January 24th, 2025
Accepted: February 05th, 2025 | unprocessed images. When compared to existing methods that rely solely on traditional image enhancement techniques, the CLAHE method does not involve image blending. CLAHE, combined with image blending, showed superior performance in highlighting disease symptoms, thereby leading to more accurate predictions. These findings highlight the crucial role of effective disease detection in tomato crops, as timely identification of health issues can lead to more informed management decisions and enhanced yield. By facilitating higher accuracy rates in disease detection, this research underscores the importance of advanced image preprocessing methods in developing robust machine learning solutions, ultimately enhancing decision- | | Keywords: | making processes in crop management and improving production efficiency. In this research, we conduct numerous experiments on | | AHE; computer vision; | various machine learning algorithms to identify and evaluate the | | CLAHE; disease detection; | algorithm that performs best in predicting diseases in tomatoes | | KNN; random forest; SVM | based on the provided image of a tomato leaf. | 1. Introduction The tomato plant is a vital crop that is consumed globally, yet it faces significant threats from various diseases that can lead to substantial yield losses. These diseases, if left undetected or untreated, can cause devastating effects on both the quantity and quality of the harvest. Despite the importance of early disease detection, significant challenges remain in accurately identifying disease symptoms at early stages, particularly due to the complex and variable nature of disease progression in tomato plants. Early detection of these diseases is essential, as timely intervention can prevent widespread damage and ensure food security. Recent advancements in machine learning and image processing offer promising alternatives for the automated detection of diseases. However, the performance of these models depends heavily on the quality of input data. Like in Arora research, where he and his team used Deep Forest to classify diseases in maize plants (Arora et al., 2020), their research suggested the implementation of Deep Forest algorithm to classify diseases present in maize plants based on their leaf images, the result of this implementation achieved an accuracy score of 96%, which demonstrates superior accuracy compared to other algorithm and model they compared to in their work. But their research relies heavily on the dataset that they used. Without an adequate preprocessing technique to further enhance the dataset's quality, the accuracy score achieved when implementing the Deep Forest algorithm can drop if some images in the dataset are poorly captur

FAQs

What preprocessing methods are proposed for tomato leaf disease detection?

The study proposes using Contrast Limited Adaptive Histogram Equalization (CLAHE) and an image blending technique to enhance tomato leaf image quality before training machine learning models.

How does CLAHE improve the tomato leaf images for disease detection?

CLAHE normalizes and improves contrast in tomato images, making key visual features more visible for more reliable analysis by machine learning algorithms.

What accuracy improvement is reported compared with baseline models?

The experimental results show improvements of up to 4.02% over baseline models that use standard unprocessed images.

Aug 3, 2026

Upload by Valentina