Research On Virtual Perspective Drawing Method Based On DIBR

With the improvement of productivity and people's pursuit of spiritual life,traditional2 D video has become increasingly unable to meet people's viewing requirements.In order to meet these rapidly growing demands,3D video gradually emerged from the tide of 2D video.In 3D video,free viewpoint video is a very important development direction,but due to the limitation of the number of viewpoints,it is not possible to place a video camera at each viewpoint.This requires that unknown viewpoints be calculated from known viewpoints.This is the virtual viewpoint technology,which is currently widely used in information processing,virtual reality and other fields.Virtual Depth Image-Based Virtual Perspective Plotting(DIBR)as a core issue in free-viewpoint video applications,the quality and speed of any video obtained from it is crucial to the development of free-viewpoint video.In order to solve the problems of overlap,voids,artifacts in the classic DIBR method,and fine cracks caused by resampling to improve the quality of the virtual viewpoint image,this paper proposes a high-quality virtual viewpoint rendering method based on improved 3D warping optimization,including improvements.3D warping optimization algorithm,improved image fusion method,and improved hole filling method.Among them,the improved 3D warping optimization algorithm starts from the entire 3D warping process,decomposes the 3D warping process into two steps and does not perform three-dimensional modeling,and proposes an improved method in two steps respectively: in the first step,through An adjustable coefficient is introduced into the transformation matrix to adjust the distortion of the three-dimensional parameter to realize the adjustment of the depth information so as to achieve the purpose of improving the image quality of the virtual viewpoint as a whole;in the second step,it mainly aims at the result of resampling.For the small crack problem,an adaptive multiple-injection algorithm was proposed to improve the cracking problem while taking into account the time complexity.The improved image fusion method achieves the quality of the virtual view image after fusion by modifying the fusion parameters.The improved hole filling method uses the preset RGB threshold to detect holes to achieve more comprehensive detection of holes,and combines the local correlation information contained in DIBR data to use the hole filling method of image search matching on the basis of comprehensive detection of holes.This method can basically detect and accurately fill most holes.The test consequence indicate that the means bring forward in this chaptercan significantly improve the quality of virtual point-of-view images.