常规图像去噪方法难以去除飞行时间技术或结构光等深度传感器所获取的深度图像中的大量噪声, 因此提出一种根据图像深度值进行分层的去噪算法。该算法首先对深度图像实现噪声强度估计, 根据噪声强度和深度图像的探测距离范围, 确立图像的深度层级间隔, 并得到分层级的深度图像, 然后对每一层级的深度图像进行去噪, 最后对分层级的图像进行拼合, 合成完整深度图像。实验结果表明, 该算法能够有效去除图像中不确定性干扰噪声, 同时较好地保持图像中目标和景物背景的原始细节以及边缘信息。

Conventional denoising methods cannot deal with much noise in depth images, which are acquired by time-of-flight or structured light depth sensors. So, a laminar denoising algorithm is proposed in this paper, which evaluates the noise level of the image, and depth image laminas are acquired with established depth laminar level according to noise level. Then, noises and holes are removed in every lamina. At last, all laminar images are combined into a whole depth image. Experimental results demonstrate that the proposed algorithm can not only remove the uncertain interferential noise, but also maintain the original detail and edges of objects in the depth images.