| 保留边界特征的深度图像增强算法研究 |
| |
| 引用本文： | 周自顾,曹杰,郝群,高泽东,肖宇晴.保留边界特征的深度图像增强算法研究[J].应用光学,2018,39(2):200-206. |
| |
| 作者姓名： | 周自顾 曹杰 郝群 高泽东 肖宇晴 |
| |
| 作者单位： | 1.北京理工大学 光电学院，机器人与系统教育部重点实验室 北京 100081 |
| |
| 基金项目： | 国家自然科学基金61275003国家自然科学基金91420203国家自然科学基金61605008国家重大科学仪器设备开发专项项目2014YQ350461深圳市基础研究(学科布局)项目JCYJ20170412171011187 |
| |
| 摘 要： | 针对现有深度图像增强算法存在边界保留特性差的问题，提出梯度掩模导向联合滤波(gradient mask guided joint filter, GMGJF)算法。利用深度图像进行Sobel梯度变换获取边界方向信息，利用深度图像空洞区域生成空洞掩模，再以边界方向和空洞掩模为导向联合彩色图像对深度图像进行迭代高斯滤波和空洞填充。实验结果表明，GMGJF算法的PSNR(peak signal to noise ratio)、SSIM(structural similarity index measure)比IMF(iterative median filter)、GF(guided filter)、JBF(joint bilateral filter)算法的PSNR、SSIM至少提高了3.50%和1.07%，不仅去噪能力、空洞填充能力最强，而且边界特征保持最好，有利于深度图像的特征提取与目标识别。
|
| 关 键 词： | 深度图像 梯度掩模导向联合滤波 空洞掩模 图像增强 PSNR |
| 收稿时间： | 2017-10-19 |
Depth image enhancement algorithm for preserving boundary |
| |
| Affiliation: | 1.School of Optoelectronics, Beijing Institute of Technology, Key Laboratory of Biomimetic Robots and Systems(Ministry of Education), Beijing 100081, China2.Graduate School at Shenzhen, Tsinghua University, Shenzhen 518055, China3.Xi'an Institute of Applied Optics, Xi'an 710065, China |
| |
| Abstract: | The drawback of current depth image enhancement algorithms is poor performance of edge preserving. To solve this drawback, the gradient mask guided joint filtering(GMGJF)algorithm is proposed. The Sobel gradient transform is used to obtain the boundary direction information, and the hole region of the depth images was utilized to generate the hole mask.Furthermore, taking the boundary direction and the cavity mask as the guidance, the color image was jointed to perform iterative Gaussian filtering and hole filling on the depth image. Experimental results show that the peak signal to noise ratio(PSNR)and the structural similarity index measure(SSIM) of GMGJF algorithm are improved by at least 3.50% and 1.07% respectively, compared with the iterative median filter(IMF), guided filter (GF) and joint bilateral filter(JBF) algorithms, it has both the strongest ability of denoising and hole filling, and can remain the boundary features best, which is good for feature extraction and target recognition of depth image. |
| |
| Keywords: | |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《应用光学》浏览原始摘要信息 |
|
点击此处可从《应用光学》下载全文 |
|
