摘要
针对现有DIBR虚拟视点图像渲染算法存在的问题,提出一种基于深度信息进行后处理的DIBR虚拟视点渲染算法.该算法利用摄像机参数和深度信息,将参考视点图像的纹理和深度图变形到虚拟视点,利用基于深度信息的后处理消除变形图像中的裂缝和伪影,结合深度信息采用inpaint技术填补剩余的纹理空洞.实验结果显示,该方法不论从主观质量或是客观质量上均优于现有的虚拟视点生成方法,且计算量也不高.在数值试验中,还对多视点渲染算法的图像合成质量进行比较,对视频压缩的影响进行了分析.
This thesis presents a DIBR virtual viewpoint rendering algorithm based on the depth of information processing. The algorithm use of camera parameters and depth of information, and the reference viewpoint images and depth images are wrapped to the virtual viewpoint. Cracks and artifacts were removed by pos tprocessing the projected images based on the depth of information. Using the depth information, the remaining disocclusions were filled by inpainting techniques, and then these images are blended together. The experiments indicate that the method is superior to the existing methods of generating virtual viewpoints, not only in the subjective quality but also in the objective quality, and has low complexity. The study compares the quality of the composite image of the multi-viewpoint rendering algorithms with the old, and analyses the influence of the video compression.
出处
《湖北工业大学学报》
2013年第1期36-42,63,共8页
Journal of Hubei University of Technology
基金
国家自然科学基金资助项目(60672134
61071091
61271082)
国家自然科学基金青年科学基金项目(61001152)
湖北省教育厅科学研究(重点)项目(D20121408)
