Research On Depth Map Reconstruction And Coding Of 3DTV

High quality depth acquisition has been a challenge for both academia and industry for a long time. Existing depth sensing technologies have many shortcomings, to some extent, in terms of resolution, completeness and accuracy, which heavily limits the development of related applications such as 3DTV and machine vision. Meanwhile, the data amount of 3DTV is more than that of monoscopic video and depth information has a lot of redundancy, which bring inconvenience to the storage and transmission of 3D videos. Therefore, high quality depth acquisition and effective depth map compression are two core technologies in 3DTV field.A new depth recovery method based on geodesic distance filtering is proposed to reconstruct high quality depth maps from low quality ones. First, we introduce the definition of geodesic distance and improve it by adding depth guidedmap to restrict depth edges. Then we calculate the geodesic distance transform according to the improved geodesic distance with the high quality color image and the low quality depth image. Last, the recovered depth image is obtained by using the multilateral filter with the transform results. In addition, the algorithm is accelerated so that we can achieve real time performance. We also use the CUDA parallel computation to accelerate the algorithm, making depth recovery more fast. We compare our method with state-of-the-art schemes on the benchmark dataset and real systems, and our approach out-performs existing state-of-the-art schemes.For the depth compression, we apply our depth reconstruction method to the depth compression based on depth down/up sampling compression method, which can effectively compress the depth maps and get good effect depth codec result.