基于图像边缘特征的深度上采样算法

针对低分辨率深度图像上采样容易导致边缘模糊问 题,提出了一种基 于图像边缘特征的深度上采样算法。一方面,利用相同低分辨率深度和彩色图像的相关性系 数,自适应调 节深度图像边缘上采样过程中深度和彩色的权重；另一方面,结合上采样值和低分辨率深度 图像中邻近像 素值,对低分辨率深度图像的不连续区域进行求精操作以进一步减少边缘模糊现象。实验结 果表明,本文算法 的性能优于近年文献中提出的算法。本算法上采样深度图像的平均坏点率(B PR)为2.07%,均方根误差(RMSE )为3.46,峰值信噪比(PSNR)为38.58dB,绘制 虚拟视点的平均PSNR为39.58dB。

Depth upsampling algorithm based on image edge characteristics

Depth camera based on the principle of time of flight can capture dept h maps in real time. Because the resolution of the captured depth maps is lower than that of the corr esponding color images, they need to be upsampled so as to represent the depth perception information of scenes reasonably.To address the problem that the upsampling methods for low resolution depth images are prone to result in edge blurring,a depth upsampling algorithm based on image edge characteristics is proposed.On the one hand,the correlation coefficient of depth and color images of the same low resolution are used to adjust the weight of depth and color in the process of depth image edge upsampli ng adaptively.On the other hand,discontinuous regions in the low resolution depth images are refined to further reduce the edge blurring phenomenon by combining the upsampled value with neighbor pixel va lues of low resolution depth image.Experimental results show that the proposed algorithm ou tperforms other algorithms reported in the literatures.The average bad pixel rate,root mean sq uare error and peak signal to noise ratio (PSNR) of the upsampled depth imag es are 2.07%,3.46and 38.58dB,respectively. The average PSNR of the rendered virtual view is 39.58d B.