Multiscale Retinex Infrared Image Enhancement Based on the Fusion of Guided Filtering and Logarithmic Transformation Algorithm
-
摘要: 针对采用红外成像仪获取红外图像边缘模糊、对比度差等缺点造成图像视觉效果差、质量低等问题。以多尺度Retinex算法为框架,依据引导滤波保边和梯度保持性,提出引导滤波和对数变换算法融合的多尺度Retinex红外图像增强方法。首先,用引导滤波替换MSR算法中的高斯滤波来估计照度分量。其次,将照度分量经过对数变换处理,执行低灰度部分扩展和高灰度部分压缩。最后,引导滤波分割得到的细节层图像线性放大并与MSR(多尺度Retinex)处理后的图像叠加,获得增强的红外图像。实验证明,与传统MSR算法和引导滤波相比该算法效果明显,可以有效地提高红外图像质量。
-
关键词:
- 红外图像 /
- 图像增强 /
- 多尺度Retinex /
- 引导滤波 /
- 对数变换
Abstract: Problems such as blurred edges and poor contrast in infrared images acquired by an infrared imager lead to poor visual effects and low image quality. Based on the multi-scale Retinex (MSR) algorithm, a MSR infrared image enhancement method based using guided filter edge preserving and gradient preserving is proposed. Firstly, a guided filter is used in place of the Gaussian filter in the MSR algorithm to estimate the illuminance component. Secondly, the illumination component is processed via logarithmic transformation, expanding the low end of the gray scale and compressing the high end. Finally, the detail layer image obtained using guided filtering is linearly amplified and superimposed with the MSR processed image to obtain an enhanced infrared image. Experimental results demonstrate that the proposed algorithm can effectively improve the quality of infrared image compared with the conventional MSR algorithm and guided filter.-
Key words:
- infrared image /
- image enhancement /
- MSR /
- guided filtering /
- Logarithmic transformation
-
表 1 场景1客观评价结果
Table 1. Scenario 1 objective evaluation results
Algorithm Evaluation parameters Information entropy Mean gradient Peak signal-to-noise ratio Original photo 7.2768 18.1815 - MSR algorithm 7.4062 15.2745 34.0826 Guided filter 7.3250 11.8575 82.5478 Ours 7.5202 24.9135 71.7486 表 2 场景2客观评价结果
Table 2. Scenario 2 objective evaluation results
Algorithm Evaluation parameters Information entropy Mean gradient Peak signal-to-noise ratio Original photo 6.4174 5.6610 - MSR algorithm 6.0917 3.2130 10.0513 Guided filter 6.5708 2.8815 86.6363 Ours 6.6079 9.0015 66.9985 表 3 场景3客观评价结果
Table 3. Scenario 3 objective evaluation results
Algorithm Evaluation parameters Information entropy Mean gradient Peak signal-to-noise ratio Original photo 6.7321 11.7166 - MSR algorithm 6.8226 10.0725 14.1633 Guided filter 6.4210 6.7830 82.9106 Ours 7.2111 17.9520 68.1152 -
[1] 莫春和, 段锦, 付强, 等. 国外偏振成像军事应用的研究进展(下)[J]. 红外技术, 2014, 36(4): 265-270. http://hwjs.nvir.cn/article/id/hwjs201404002MO Chunhe, DUAN Jin, FU Qiang, et al. Review of polarization imaging technology for international military application(II)[J]. Infrared Technology, 2014, 36(4): 265-270. http://hwjs.nvir.cn/article/id/hwjs201404002 [2] 龚璐, 沈国土, 蔡继光, 等. 典型物体表面温度的测量及初步计算[J]. 红外技术, 2013, 35(7): 439-442. http://hwjs.nvir.cn/article/id/hwjs201307011GONG Lu, SHEN Guotu, CAI Jiguang, et al. Measurement and preliminary calculation of typical object surface temperature[J]. Infrared Technology, 2013, 35(7): 439-442. http://hwjs.nvir.cn/article/id/hwjs201307011 [3] 万智萍. 结合视觉特性与灰度拉伸的直方图均衡化红外图像算法[J]. 计算机工程与设计, 2016, 37(3): 714-719. https://www.cnki.com.cn/Article/CJFDTOTAL-SJSJ201603030.htmWAN Zhiping. Histogram equalization infrared image algorithm combining visual characteristics and gray stretching[J]. Computer Engineering and Design, 2016, 37(3): 714-719. https://www.cnki.com.cn/Article/CJFDTOTAL-SJSJ201603030.htm [4] Land E H, Mccann J J. Lightness and Retinex theory[J]. Journal of the Optical Society of America, 1971, 61(1): 1-11. doi: 10.1364/JOSA.61.000001 [5] 梁栋, 顾杰宁, 张陈, 等. 基于小波变换的红外热成像图像处理的无损检测技术[J]. 物联网技术, 2020, 10(3): 37-39. https://www.cnki.com.cn/Article/CJFDTOTAL-WLWJ202003016.htmLIANG Dong, GU Jiening, ZHANG Dong, et al. Nondestructive testing technology of infrared thermal imaging image processing based on wavelet transform[J]. Internet of Things Technology, 2020, 10(3): 37-39. https://www.cnki.com.cn/Article/CJFDTOTAL-WLWJ202003016.htm [6] Jobson Daniel J, Rahman Zia-ur, Woodell Glenn A. Properties and performance of a center surround Retinex[J]. IEEE Transactions on Image Processing, 1997, 6(3): 451-462. doi: 10.1109/83.557356 [7] Jobson Daniel J, Rahman Zia-ur, Woodell Glenn A. A multi-scale Retinex for bridging the gap between color images and the human observation of scenes[J]. IEEE Transactions on Image Processing, 1997, 6(7): 965-976. doi: 10.1109/83.597272 [8] 温海滨. 基于局部正则先验的Retinex红外图像细节增强方法研究[D]. 西安: 空军工程大学, 2017.WEN Haibing. Research on Detail Enhancement Method of Retinex Infrared Image Based on Local Regularization a Priori[D]. Xi'an: Air Force Engineering University, 2017. [9] 任崇巍. 基于双边滤波和Retinex算法的货车图像预处理方法[J]. 现代城市轨道交通, 2019(6): 39-43. https://www.cnki.com.cn/Article/CJFDTOTAL-XDGD201906007.htmREN Congwei. Freight car image preprocessing method based on bilateral filtering and Retinex algorithm[J]. Modern Urban Transit, 2019(6): 39-43. https://www.cnki.com.cn/Article/CJFDTOTAL-XDGD201906007.htm [10] 常戬, 任营, 贺春泽. 改进双边滤波Retinex的多聚焦图像融合[J]. 中国图象图形学报, 2020, 25(3): 432-441. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGTB202003002.htmCHANG Jian, REN Ying, HE Chunze. Multi focus image fusion based on improved bilateral filtering Retinex[J]. Journal of Image and Graphics, 2020, 25(3): 432-441. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGTB202003002.htm [11] 王秋云, 杨昊亮, 李雄伟, 等. 基于Lab色彩空间改进的多尺度Retinex图像增强算法[J]. 中国刑警学院学报, 2020(1): 123-128. https://www.cnki.com.cn/Article/CJFDTOTAL-XING202001017.htmWANG Qiuyun, YANG Haoliang, LI Xiongwei, et al. Improved multi-scale Retinex image enhancement algorithm based on Lab color space[J]. Journal of Criminal Investigation Police University of China, 2020(1): 123-128. https://www.cnki.com.cn/Article/CJFDTOTAL-XING202001017.htm [12] 杨福豪, 史启超, 蓝方鸣, 等. 基于色彩衰减补偿和Retinex的水下图像增强[J]. 宁波大学学报: 理工版, 2020, 33(1): 58-64. https://www.cnki.com.cn/Article/CJFDTOTAL-NBDZ202001009.htmYANG Fuhao, SHI Qichao, LAN Fangming, et al. Underwater image enhancement based on Retinex [J]. Journal of Ningbo University: Natural Science and Engineering, 2020, 33(1): 58-64. https://www.cnki.com.cn/Article/CJFDTOTAL-NBDZ202001009.htm [13] Roberts J W, Aardt J A V. Assessment of image fusion procedures using entropy, image quality, and multispectral classification[J]. Journal of Applied Remote Sensing, 2008, 2(1): 1-28.