DEPTH ESTIMATION METHOD BASED ON SPARSE AD-CENSUS AND ON LIGHT FIELD OCCLUSION MODEL

• Main Authors: LU, Guanghui, MA, Yun, WANG, Litong, CAI, Hongbin, SHUI, Wenli, ZHAO, Chen
• Format: Patent
• Language: Chinese; English; French
• Published: 06.06.2024
• Subjects: CALCULATING; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

The present invention provides a depth estimation method based on sparse AD-Census and on a light field occlusion model. The method comprises: inputting a light field image, calculating a multi-view image, calculating a stereo matching-based cost value C stereo, calculating a cost value C occ based on a light field occlusion model, and globally optimizing a disparity map on the basis of graph cut algorithm. In order to obtain a matching cost having higher accuracy, the method introduces AD-Census having higher robustness into a light field stereo matching method, and reduces the amount of calculation of the cost values by means of sparse Census and cross viewing points. In addition, in order to reduce the depth estimation error of the light field depth estimation method at an object edge area due to the occlusion problem, the method improves a global information-based energy function for acquiring a disparity map. Specifically, on the one hand, the method subjects the cost value C stereo and the cost value C