Parallelly processable multi-view image synthesis method in imaging technology

Abstract

The invention discloses a parallelly processable multi-view image synthesis method, which comprises the following steps: 1) capturing reference images to be synthesized into multi-view images, and defining the two view images of the reference images respectively as an image 1 and an image 2; 2) taking the image 1 as a standard and the image 2 as a reference to generate a depth map 1; 3) taking the image 2 as the standard and the image 1 as the reference to generate a depth map 2; 4) generating N / 2-1 multi-view images based on a depth image drawing method according to the depth map 1 and the image 1; and 5) generating N / 2-1 multi-view images cooperatively based on the depth image drawing method according to the depth map 2 and the image 2. According to the technical scheme of the invention, multi-view diagrams can be obtained by the parallel processing of two groups of two-dimensional pictures and depth maps, the generation times can be reduced, the generated vacancies are reduced, and the vacancy scope is narrowed, so the workload and the time of repair are reduced, and image generation is accelerated.

Description

technical field [0001] The invention relates to the technical field of multi-view and three-dimensional image synthesis, in particular to a multi-view image synthesis method capable of parallel processing in imaging technology. Background technique [0002] Following high-definition television, stereoscopic television (3DTV) will become the development direction of the next-generation television technology. The traditional stereoscopic video is obtained by fixing the left and right cameras in the scene to simulate the eyes of people from two different angles. This results in the need to transmit two channels of video when transmitting programs. In order to reduce data transmission, both the joint video team JVT and the European ATTEST project suggest adopting the data format of "video + depth". In this way, the left and right videos can be synthesized by the middle video and its corresponding depth image (depth iamge), which is the depth image based rendering (Depth Image ...

AI Technical Summary

Problems solved by technology

[0004] Because the existing technology only uses one of the two-dimensional images and the synthesized depth map to generate images of different viewing angles, the resources of another image are wasted; and when multiple images of different viewing angles are generated using depth-based graphics rendering technology (DIBR), Only one two-dimensional image and depth image are used, which requires more generation times, and the holes generated by the generated view will increase and become larger, which requires more time and cost for image generation and repair work, and when the repair The more parts there are, the more serious the inpainted image deviates from the original image.

At the same time, the current multi-view image processing method can only be processed sequentially, and cannot be processed in parallel, and it takes a lot of time to generate views

Images