Related Experiment Video
Updated: Jan 10, 2026
05:56
Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application
Published on: April 14, 2023
3.2K
A Deep Regression Model for Tongue Image Color Correction Based on CNN
Xiyuan Cao1, Delong Zhang1, Chunyang Jin1
1State Key Laboratory of Extreme Environment Optoelectronic Dynamic Testing Technology and Instrument, North University of China, Taiyuan 030051, China.
Journal of Imaging
|November 26, 2025
Summary
TococoNet, a novel convolutional neural network, effectively corrects color bias in tongue images. This deep learning model enhances diagnostic accuracy by restoring true colors in medical imaging.
Area of Science:
- Medical Imaging
- Computer Vision
- Artificial Intelligence
Background:
- Image color authenticity is compromised by varying viewing conditions, leading to visual inconsistencies.
- Deep learning offers advanced solutions for image processing and optimization challenges.
Purpose of the Study:
- To introduce TococoNet (Tongue Color Correction Network), a novel regression model for eliminating color bias in tongue images.
- To evaluate TococoNet's effectiveness in color correction for diagnostic applications.
Main Methods:
- Developed a convolutional neural network (CNN) model, TococoNet, featuring symmetric encoder-decoder U-Blocks and M-Block for feature fusion.
- Trained the model using simulated common biased colors and conducted correction experiments with random color biases.
- Utilized image quality indicators and color distance (ΔE) to assess correction performance.
Main Results:
- TococoNet demonstrated superior color correction compared to conventional algorithms and shallow networks.
- Achieved up to 84% correction effectiveness (ΔE) for tongue images with random color casts.
- Significantly reduced maximum ΔE from 30.38 to 6.05 in actual captured images.
Conclusions:
- TococoNet exhibits excellent color correction capabilities for tongue images.
- The model shows promising potential for clinical assistance and automatic tongue diagnosis.
