Related Experiment Video
Updated: Jan 12, 2026
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
1.0K
Contrast limited adaptive histogram equalization (CLAHE) and colour difference histogram (CDH) feature merging
Steve Okyere-Gyamfi1,2, Michael Asante1, Kwame Ofosuhene Peasah1
1Department of Computer Science, Kwame Nkrumah University of Science and Technology, Kumasi, Ghana.
Plos One
|October 31, 2025
Summary
A new deep learning model, CCFM-CapsNet, effectively detects plant leaf diseases with high accuracy, outperforming traditional methods. This advancement aids in crop yield enhancement and supports Sustainable Development Goal 2 (Zero Hunger).
Area of Science:
- Computer Vision
- Deep Learning
- Agricultural Technology
Background:
- Accurate leaf disease detection is vital for crop yield and food security.
- Traditional Convolutional Neural Networks (CNNs) face challenges with image variations and require extensive data.
- Capsule Networks (CapsNets) offer an alternative but have limitations in handling complex images.
Purpose of the Study:
- To introduce a novel Capsule Network (CapsNet) model, CCFM-CapsNet, designed to overcome the limitations of existing deep learning architectures for plant disease detection.
- To improve the accuracy and efficiency of identifying plant leaf diseases using advanced deep learning techniques.
- To evaluate the performance of CCFM-CapsNet on diverse plant disease datasets and benchmark datasets.
Main Methods:
- Developed CCFM-CapsNet by integrating Contrast Limited Adaptive Histogram Equalization (CLAHE) for noise reduction and a novel feature extraction method (CDH).
- Incorporated max-pooling and dropout layers into the original CapsNet architecture.
- Tested the model on datasets including various plant leaf diseases (apples, bananas, grapes, etc.) and standard image classification datasets (Fashion-MNIST, CIFAR-10).
Main Results:
- CCFM-CapsNet achieved high validation accuracies across multiple datasets, including 99.53% for apples, 100% for corn, and 100% for rice.
- The model demonstrated superior performance compared to traditional CapsNet and other advanced CapsNet variants.
- Achieved these results with a relatively small number of parameters (millions).
Conclusions:
- CCFM-CapsNet is a robust and accurate model for identifying plant leaf diseases, offering a significant improvement over existing methods.
- The model's efficiency and high accuracy make it a valuable tool for smart agriculture and achieving Sustainable Development Goal 2 (Zero Hunger).
- The integration of CLAHE and CDH effectively enhances feature extraction and noise reduction in complex image recognition tasks.
