Depth Map Upsampling Method Based on Shortest Joint Path

Abstract

The invention gives a depth image up-sampling method based on a shortest combined path, and mainly solves the problem of depth edge blur and depth blending in an up-sampling depth image obtained in the prior art. The depth map up-sampling method comprises the following implementation steps: 1) inputting a high-resolution color image and a corresponding low-resolution depth image, causing the low-resolution depth image to correspond to a high-resolution grid to obtain a high-resolution depth image to be filled; 2) calculating four components including Euclidean distance, color difference, gradient difference and boundary, and carrying out weighting on the four components to define the shortest combined path; 3) adopting a global traversal search algorithm to find a seed point which has the shortest combined path to each target point; and 4) endowing the corresponding target point with the depth value of the searched seed point to finish the depth image up-sampling so as to obtain the high-resolution depth image. By use of the depth map up-sampling method, the up-sampling depth image with a clear edge can be quickly and efficiently generated, an up-sampling result is stable under different amplification multiples, and the depth map up-sampling method can be used for a 2D (Two-dimensional) to 3D (Three-dimensional) technology and stereotelevisions.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a depth image upsampling method, which can be used for converting 2D to 3D and stereoscopic television. Background technique [0002] Because 3D video can provide an immersive virtual environment, it is considered to be the main development trend of future multimedia. Depth plays an important role in 3D applications such as human-computer interaction and 3D reconstruction, and the real-time capture of depth images has received extensive attention. However, the production cost of the depth sensor is high, and the improvement rate of its resolution is far lower than the market demand, so it is necessary to obtain a high-resolution depth map through some 3D technologies. The traditional methods mainly include laser scanning method and passive stereo vision method. The former has high requirements on equipment and cannot achieve real-time capture. The latter can...

AI Technical Summary

Problems solved by technology

[0007] Although the above-mentioned algorithms can obtain better upsampling results in homogeneous areas, they cannot make full use of color image information and the information transmission is not global, resulting in unclear edges of high-resolution depth images.

Images