光学学报, 2015, 35 (3): 0310002, 网络出版: 2015-02-04
基于多尺度Retinex的非下采样Contourlet域图像增强
Image Enhancement in Non-Subsampled Contourlet Transform Domain Based on Multi-Scale Retinex
遥感 高光谱图像 图像增强 非下采样Contourlet变换 多尺度Retinex 混沌小生境粒子群优化 remote sensing hyperspectral image image enhancement non-subsampled contourlet transform multi-scale Retinex niche chaotic particle swarm optimization
摘要
针对部分遥感图像和高光谱图像中存在的对比度不足、整体偏暗等问题,提出了一种基于多尺度Retinex(MSR)和混沌小生境粒子群优化(NCPSO)的非下采样Contourlet变换(NSCT)域图像增强方法,用于改善图像质量。对图像进行NSCT 分解,得到一个低频分量和多个不同方向的高频分量;在低频分量上进行混合灰度函数的多尺度Retinex增强;同时利用非线性增益函数调整高频分量系数,将兼顾对比度和信息熵的定量综合评价函数作为NCP-SO 的适应度,寻找非线性增益函数所涉及的最优参数。大量实验结果表明,与双向直方图均衡方法、NSCT方法、多尺度Retinex 方法、平稳小波变换和Retinex 方法等4 种增强方法相比,提出的方法能更有效地提高图像的对比度和信息熵,增强图像的整体视觉效果。
Abstract
Aiming at the problem of contrast deficiency and low luminance in some remote sensing images and hyperspectral images, an enhancement method in non-subsampled contourlet transform (NSCT) domain based on multi-scale Retinex (MSR) and niche chaotic mutation particle swarm optimization (NCPSO) is proposed to improve the quality of images. Firstly, an image is decomposed through NSCT. A low-frequency component and several high-frequency components in different directions are produced. Then the low-frequency component is enhanced by the multi-scale Retinex algorithm with hybrid intensity transfer function. While the coefficients of high-frequency components are adjusted to enhance the edges by nonlinear gain function. The optimal parameters in the nonlinear gain function are searched by the niche chaotic particle swarm optimization algorithm, whose fitness is the integrated quantitative evaluation function considering both contrast and information entropy. A large number of experimental results show that, compared with four enhancement methods such as histogram double equalization method, non-subsampled contourlet transform method, multi-scale Retinex method and stationary wavelet transform and Retinex method, the proposed method can improve the contrast and information entropy more efficiently, and enhances the whole visual effects.
吴一全, 史骏鹏. 基于多尺度Retinex的非下采样Contourlet域图像增强[J]. 光学学报, 2015, 35(3): 0310002. Wu Yiquan, Shi Junpeng. Image Enhancement in Non-Subsampled Contourlet Transform Domain Based on Multi-Scale Retinex[J]. Acta Optica Sinica, 2015, 35(3): 0310002.