基于小波变换的雾霾立体图像增强算法研究  

Wavelet transform stereoscopic images enhancement algorithms based on fog and haze

作　　者：邱奕敏[1] 周毅[2,3] 

机构地区：[1]武汉科技大学信息科学与工程学院,武汉430081 [2]武汉大学计算机学院,武汉430072 [3]中冶南方(武汉)信息技术工程有限公司,武汉430223

出　　处：《计算机工程与应用》2015年第9期30-33,共4页Computer Engineering and Applications

基　　金：国家自然科学基金(No.61201423;No.61105058);校青年科技骨干培育计划项目(No.2014xz011)

摘　　要：针对立体图像在雾霾环境下的质量问题,运用小波变换的多尺度特征,提出了一种雾霾环境下的立体图像增强算法,主要用于中度污染情况下的雾霾立体图像,以提高图像资源的清晰程度。该算法将原始雾霾立体图像的深度信息与多尺度小波分解相结合,在不同尺度下分解得到的小波高频子图中设置人为操控因子,调控对比度增强的强度;锐化分解后的小波低频子图边缘来突出整体轮廓。实验从PSNR指标、视觉效果和DMOS主观评价值三个方面验证了算法的成效,该方法的增强性能均好于传统的边缘锐化和四层小波变换方法,具备很好的图像边缘增强能力,细节保护能力,且与传统小波变换有相同的算法时间复杂度。This paper proposes a new image enhancement algorithm based on edge sharpening of wavelet coefficients for fog and haze stereoscopic images, using multi-scale characteristic of wavelet transform in order to improve the clarity of fog and haze stereoscopic images, which is mainly used in moderate pollution. The algorithm combines the depth of stereo-scopic images with multi-scale wavelet decomposition, setting a control factor in the high-frequency sub-graph to regulate contrast enhancement. And it highlights the overall outline through the sharpening of the low-frequency sub-graph. Experi-mental results show that whether PSNR or visual effect, or the subjective assessment of the DMOS value, the proposed method has better enhanced performance than the conventional edge sharpening and wavelet transform. And it has good image edge enhancement, details protection. Meanwhile, the proposed algorithm has the same computational complexity with wavelet transform.

关 键 词：立体图像 小波系数 图像增强 质量评估 雾霾 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]