Machine Vision and Applications ( IF 3.3 ) Pub Date : 2021-03-23 , DOI: 10.1007/s00138-021-01189-3 Yue Zhang , Zhenfang Liu , Min Huang , Qibing Zhu , Bao Yang
Depth degradation caused by the conditions and environment of depth sensor hardware restricts its application potential, and this limitation cannot be avoided simply by improving the design of sensor. To overcome this limitation, we propose a multi-resolution depth image restoration method. Firstly, the sub-images of depth image and color image at different scales are obtained by multi-resolution analysis based on two-dimensional discrete wavelet transform. The multi-resolution joint bilateral filtering is then applied to the approximation low-frequency sub-image of the decomposed image. At the same time, using color-guided filtering method to restore high-frequency sub-images can effectively suppress edge artifacts without adding extra time burden. The high-quality output image is finally reconstructed using two-dimensional inverse discrete wavelet transform. A color guide image with rich edge information is introduced into the depth sub-image restoration to improve the depth image edge detail. Extensive experiments with synthetic and real datasets demonstrate that the proposed algorithm can effectively reduce additive Gaussian noise without losing sharp details in the noisy images and reduce the time consumption of depth image restoration.
中文翻译:
多分辨率深度图像恢复
由深度传感器硬件的条件和环境引起的深度退化限制了其应用潜力,而仅通过改进传感器的设计就无法避免这种局限性。为了克服此限制,我们提出了一种多分辨率深度图像恢复方法。首先,基于二维离散小波变换,通过多分辨率分析获得深度图像和彩色图像不同比例的子图像。然后将多分辨率联合双边滤波应用于分解图像的近似低频子图像。同时,使用色彩导引滤波方法来还原高频子图像可以有效地抑制边缘伪影,而不会增加额外的时间负担。最终,使用二维逆离散小波变换重建了高质量的输出图像。将具有丰富边缘信息的彩色导引图像引入深度子图像恢复中,以改善深度图像边缘细节。大量的合成和真实数据集实验表明,该算法可以有效地减少加性高斯噪声,而不会在噪点图像中丢失清晰的细节,并减少了深度图像恢复的时间消耗。