作者：唐露新¹, 张宇维¹, 宋有聚², 王小桂³, 于丽敏¹

作者单位：1. 广东工业大学机电工程学院, 广东 广州 510006;
2. 深圳市施罗德工业测控设备有限公司, 广东 深圳 518055;
3. 深圳市一零科技服务有限公司, 广东 深圳 518055

关键词：低纹理;形态学重建;小波变换;高光修复

摘要：

高光区域多出现于低纹理材料的平滑表面上，严重影响图像采集处理效果。针对现有图像去高光技术中过分依赖特定对象纹理特征的问题，该文提出一种利用形态学重建检测高光区域、利用小波变换修复高光区域低纹理图像序列的高光检测与抑制方法。对地下混凝土、金属等管道内壁表面高光区域进行修复，取得较好效果；通过修复胶合板等多种低纹理经典材质，对比其他方法，并应用差分图像的均值、方差与互相关系数评价修复效果。结果表明，修复后的多种材质与原图相似度在0.87以上，比其他方法的相似度平均提高9.7%与6.7%，且可同时应用于多种材料的高光区域处理，具有较好的应用前景。

Highlight of low texture image sequences restoration based on morphological reconstruction and wavelet transform
TANG Luxin¹, ZHANG Yuwei¹, SONG Youju², WANG Xiaogui³, YU Limin¹
1. Guangdong University of Technology, Guangdong, Guangzhou 510006, China;
2. Shenzhen Srod Robotics Industrial Measurement and Control Equipment Co., Ltd., Shenzhen 518055, China;
3. Shenzhen Srod One Zero Science and Technology Service Co., Ltd., Shenzhen 518055, China
Abstract: Highlight region is mostly on the smooth surface with low texture, which seriously affects the effect of image acquisition. In view of the problem that the existing image sequence is overly dependent on the texture matching in highlight technology, a method of high light detection and suppression by using morphological reconstruction to detect highlight region and wavelet transform to repair low texture image sequence is proposed. The highlight area of the inner surface of pipes, such as concrete and metal, was repaired. At the same time, comparative experiments were carried out on various low texture materials, such as plywood. The mean, variance and correlation coefficient of difference images were used to evaluate the restoration. Results show that the similarity of different materials restored by this method is above 0.87, and the average increase of 9.7% and 6.7%.
Keywords: low texture;morphological reconstruction;wavelet transform;highlight restoration
2019, 45(4):109-115  收稿日期: 2018-10-20;收到修改稿日期: 2018-11-15
基金项目:
作者简介: 唐露新(1958-),男,湖南冷水江市人,教授,硕士,研究方向为图像检测与信号处理、传感器网络与测控、无人机技术与定位、精密仪器与测控系统、舞台灯光技术与控制、机电一体化技术
参考文献
[1] 李波锋. 基于机器视觉的排水管道缺陷检测算法研究[D]. 广州:广东工业大学, 2015.
[2] KIM H, JIN H, HADAP S, et al. Specular reflection sepa-ration using dark channel prior[C]//2013 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IE-EE, 2013.
[3] NGUYEN T, NHAT V Q, KIM S H, et al. A novel and effective method for specular detection and removal by tensor voting[C]//IEEE International Conference on Image Processing. IEEE, 2015.
[4] 柴玉亭, 王昭, 高建民, 等. 基于频域滤波的高光去除方法[J]. 激光与光电子学进展, 2013, 50(5):135-142
[5] 尹芳, 陈田田, 吴锐, 等. 均场退火算法在单幅灰度图像高光检测与恢复中的应用[J]. 计算机辅助设计与图形学学报, 2017, 29(5):829-837
[6] ANTONIO C S SOUZA, MARCIO C F MACEDO, VERONICA P. NASCIMENTO, et al. Real-Time High-quality specular highlig-ht removal using efficient pixel clustering[C]//Patterns and Images(SIBGRAPI), 2018 SIBGRAPI Conference on Gra-phics, 2018.
[7] PARTHA P B, RAPPY S H, KIDOO K. HDR ima-ge from single LDR image after removing highlight[C]//IEEE International Conference on Consumer Electronics(ICCE). IEEE, 2018.
[8] 汪铖杰. 基于多视角的去高光技术及应用[D]. 上海:上海交通大学, 2014.
[9] 何嘉林, 唐露新, 林永强. 基于融合技术的图像去高光方法[J]. 科学技术创新, 2018(6):90-92
[10] SAMAR M. ALSALEH, ANGELICA I, et al ReTouchImage:Fusioning from-local-to-global context detec-tion and graph data structures for fully-automatic specular reflection removal for endoscopic image[J]. Computerized Medical Image and Graphics, 2019, 73(4):39-48
[11] 黄远程, 宋博文. 形态学重建与Canny结合实现机场跑道边界检测[J]. 遥感信息, 2016, 31(6):75-82
[12] 寇万里, 车嵘, 严丽娜. 基于形态学重建和边界融合的视频对象分割方法研究[J]. 通信技术, 2018, 51(4):825-828
[13] 刘娇娇. 基于小波变换和Harris算子的数字图像修复研究[D]. 大连:大连理工大学, 2014.
[14] REHAM G, ELLA H A, HASSAN E B A, et al. Multi-spectral and panchromatic image fusion approach using stationary wavelet transform and swarm flower pollination optimization for remote sensing applications[J]. Future Generation Computer Systems, 2018, 22(6):501-511
[15] 汤一平, 鲁少辉, 吴挺, 等. 基于主动式全景视觉的管道形貌缺陷检测系统[J]. 红外与激光工程, 2016, 45(11):183-189
[16] 邹玮刚, 周志辉, 王洋. 基于非降采样轮廓波变换的图像修复算法[J]. 计算机应用, 2017, 37(2):553-558