Parallax generation method, generation cell and three-dimensional video generation method and device

Abstract

Embodiment of the invention provides parallax generation method, generation unit and three-dimensional video generation method and device. The parallax generation method comprises: calculating the parallax of non-matched pixels according to a first view and a second view, and occlusion pictures of the first and the second views to obtain the parallax pictures of the first and the second views; generating new occlusion pictures of the first and the second views according to the parallax pictures thereof; ending up the iteration if it meets the ending condition of iteration, otherwise, regarding the new occlusion pictures as an occlusion pictures continuous iteration process; obtaining an optimal index according to the occlusion pictures and parallax pictures, and judging whether the iteration ends up according to the optimal index. Embodiment of the invention has implemented to select needed video images from several ones according to current view points, capable of synthesizing two or several video images to generate a three-dimensional video. In the iteration process, it only needs to update values of non-matched pixels so as to reduce the calculation time and speed up the velocity of convergence.

Description

technical field [0001] The invention relates to a three-dimensional image processing technology, in particular to a parallax generation method, a generation unit, a three-dimensional video generation method and a generation device. Background technique [0002] Now there is a new form of multimedia, 3D video. Three-dimensional video allows users to feel as if they are actually there when watching a video. [0003] The 3D video reconstruction based on the binocular vision model (two-view geometry) is a current research hotspot. By imitating the visual system of most creatures in nature, the same scene is shot from two different angles (simulating the sight seen by the two eyes of the creature) , find and match the feature point pairs of the two images (the position of the feature point in the first image in the second image, simulating the intelligent mixing in the biological vision system), and finally use the method of computational geometry to calculate each pair of match...

AI Technical Summary

Problems solved by technology

Among the existing disparity generation methods, the global matching method is better than the local matching method, but the global matching method will update all pixels in each iteration, the algorithm complexity is high, and the calculation amount is large

Images