Multi-view depth estimation method, apparatus, device, storage medium and product

Abstract

The application discloses a multi-view depth estimation method, device, equipment, storage medium and product; the method comprises the following steps: acquiring a reference view and a plurality of source views, and determining the corresponding relationship of the view angles between each source view and the reference view; generating a depth hypothesis set under the corresponding view angle of the reference view; based on the depth hypothesis set, calculating the cost volume corresponding to each source view, and fusing each cost volume to obtain a multi-view cost volume; determining a depth map under the corresponding view angle of the reference view according to the multi-view cost volume, and re-projecting the depth map to the corresponding view angle of each source view according to the corresponding relationship of the view angles between the reference view and each source view to obtain a re-projection error; determining a pixel to be corrected in the depth map according to the re-projection error, and correcting the pixel to be corrected according to a preset correction strategy to determine a target depth map. The application can efficiently determine the depth map without pre-acquiring the depth true value, and improves the accuracy and robustness of depth estimation under multi-view.

Images