基于双模式联合三边滤波器的深度图像超分辨率方法

彩色图像引导的深度图像超分辨率方法通过利用高分辨率彩色图像的高频信息来重建深度图像,取得了不错的重建效果,但当深度图像和彩色图像边缘不(完全)一致或彩色区域纹理丰富时,重建图像普遍存在边缘模糊和纹理拷贝问题.针对这一问题,提出一种边缘图像引导的双模式联合三边滤波器(DMJTF)方法.该方法利用单幅低分辨率深度图像构建了一个边缘图像金字塔字典,然后利用MRF模型构建了一个高分辨率边缘图像,该图像确定了滤波器的两种模式,分别用于重构深度图像的边缘和平滑区域.实验结果表明DMJTF算法有效地避免了纹理拷贝异常,降低了边缘模糊现象,在定性和定量两个方面都优于其他算法,取得了较好的超分效果.

Depth map super resolution based on dual mode joint trilateral filter

Color-guided depth super resolution is one way for improving the quality or resolution of the low resolution depth image, which utilizes an additional aligned high-resolution color image as assistance, however, if the smooth regions have rich color texture or the color edges do not align well with the depth edges, artifacts including texture copying and edge blurring may occur. In view of these artifacts, the paper proposed a novel dual mode joint trilateral filter(DMJTF) method for color-guided depth map super resolution. This method built an edge image pyramid dictionary by a single low-resolution depth image, and then got a high-resolution edge image by constructing a MRF model. It used the edge map as a guidance to decide when to change the filter mode, one mode was for the pixels on the edges, and the other was for the pixels in the smooth regions. Experimental results in both qualitative and quantitative aspects demonstrate the effectiveness of our approach over prior depth map upsampling works, which alleviates artifacts introduced by color image and preserves sharp edges well.