| 引用本文: | 李 季1,曾 榛1,龚洋钢2.基于改进多尺度 Retinex 和 NSST 的医学图像增强算法(J/M/D/N,J:杂志，M：书，D：论文，N：报纸).期刊名称,2026，43（1）：11-19 |
| CHEN X. Adap tive slidingmode contr ol for discrete2ti me multi2inputmulti2 out put systems[ J ]. Aut omatica, 2006, 42(6): 4272-435 |
|
| 摘要: |
| :目的 解决医学图像对比度低,边缘不清,细节不突出等问题,提高医学图像的视觉效果,保持良好的边缘信
息。 方法 提出基于改进多尺度 Retinex 和 NSST 的医学图像增强算法。 首先利用梯度引导滤波改进多尺度
Retinex,结合局部图像梯度信息,引导图像调整滤波过程,根据多尺度 Retinex 原理计算得到反射光分量及增强图;
同时,引入非下采样剪切波变换(Nonsubsampled Shearlet Transform,NSST)将图像分解得到高、低频分量,低频分量
利用梯度引导滤波处理,提取图像的结构和特征信息,提高背景对比度,通过 NSST 逆变换恢复图像,将两图利用拉
普拉斯高斯金字塔融合,增强共同特征,避免过度曝光;最后利用限制对比度自适应直方图均衡化改善局部对比度
得到最终的增强图。 结果 仿真实验和消融实验结果表明:本算法对医学图像能有效提高对比度,完整保留细节及
边缘信息。 结论 通过峰值信噪比(PSNR)、信息熵(Entropy)、基于感知的非参考图像质量评估器(PIQE)以及均值
4 项图像评价指标的客观分析,得到本算法最优,能有效改善医学图像质量,为诊断提供技术支持。 |
| 关键词: 图像增强 多尺度 Retinex NSST 拉普拉斯金字塔融合 |
| DOI： |
| 分类号: |
| 基金项目: |
|
| Medical Image Enhancement Algorithm Based on Improved Multi-scale Retinex and NSST |
|
LI Ji1，ZENG Zhen1，GONG Yanggang2
|
|
1. School of Mathematics and Statistics Chongqing Technology and Business University Chongqing 400064 China
2. School of Biomedical Engineering Southern Medical University Guangzhou 510080 China
|
| Abstract: |
| Objective This study aims to solve problems such as low contrast unclear edges and unremarkable details in
medical images improve the visual effect of medical images and preserve good edge information. Methods A medical
image enhancement algorithm based on improved multi-scale Retinex and Non-subsampled Shearlet Transform NSST is
proposed. First the method improved the multi-scale Retinex by employing gradient-guided filtering which incorporated
local image gradient information to steer the filtering process. According to the principle of multi-scale Retinex a reflected
light component was calculated and an enhanced image was generated. Meanwhile the Non-subsampled Shearlet Transform
NSST was introduced to decompose the image into high-frequency and low-frequency components. The low-frequency
components were processed by gradient-guided filtering to extract the structural and characteristic information of the image
and improve the background contrast. The image was restored through the inverse transform of NSST. The two images were
fused using the Laplacian-Gaussian pyramid to enhance the common features and avoid over-exposure. Finally the Contrast
Limited Adaptive Histogram Equalization CLAHE was used to improve the local contrast and obtain the final enhanced
image. Results The results of simulation experiments and ablation experiments showed that this algorithm could effectively
improve the contrast of medical images and completely preserve the details and edge information. Conclusion Through the objective analysis of four image evaluation indicators namely the peak signal-to-noise ratio PSNR entropy perceptual
image quality evaluator PIQE and mean M it is found that this algorithm is optimal can effectively improve the
quality of medical images and provide technical support for diagnosis. |
| Key words: image enhancement multi-scale Retinex NSST Laplacian Pyramid fusion |
