Super-resolution Reconstruction Algorithm Based On The TOF Depth Map

Depth information plays an important role in the fields of computer vision and image processing, and has broad application prospects and great economic value in the fields, such as three-dimensional visualization, three-dimensional object reconstruction, medical and biological fields, film and television production. The TOF camera can get the depth map in real-time, high frame rate, and without scanning. According to the images the TOF depth camera got have low resolution, we realize super-resolution reconstruction with the depth images.The main work of this thesis is illustrated as follows:(1)We make an introduction about the basic measurement principle, the imaging model and calibration of the TOF camera. And have an overview of the depth map degradation model and the research status and image reconstruction methods of the depth map.(2) Based on the registration of the color image and depth map from the TOF camera, we realize a depth map super-resolution method based on the geodesic distance and the second order differential operator to process the condition that the pixels have similar color but different depth value with the depth kernel function, and suppress the artifacts in the regions. Joint geodesic distance method reflects the consistency of color information and depth information. We can recover the high resolution depth map with significant edge contour. The experimental results demonstrate that the proposed approach can well preserve the edge information and solve artifacts problem, and get a depth image with high resolution.(3) Making the use of multiple TOF depth maps to achieve the depth map reconstruction method based on the maximum posterior probability. In accordance with the depth image degradation model, we use the first order norm data items which has better results to non-Gaussian noise distribution model and bilateral regularization operator which can maintain the image edge information. Then we construct objective cost function to achieve iterative optimization objective function and obtain high-resolution depth map. We can realize the reconstruction of natural objects in the scene with regular objects, irregular objects and the objects with non-smooth surface.