-
公开(公告)号:US09456193B2
公开(公告)日:2016-09-27
申请号:US14575091
申请日:2014-12-18
Applicant: Huawei Technologies Co., Ltd.
Inventor: Jing Xu , Zhimin Xu , Edmund Lam
CPC classification number: H04N13/15 , G06T3/4053 , G06T5/50 , G06T2200/21 , G06T2207/10052 , H04N5/23229 , H04N13/25 , H04N2013/0077
Abstract: The present invention provides a method and an apparatus for processing a light-field image. The method includes: acquiring a light-field image; acquiring an occlusion tag function of each unit image in the light-field image; determining a cost function for offset estimation of each unit image according to intensity change information and gradient change information that are of each unit image in the light-field image and the occlusion tag function of each unit image; acquiring an offset of each unit image in the light-field image by using the cost function; and reconstructing a high-resolution light field by using the offset of each unit image. An offset of each unit image in a light-field image may first be acquired according to optical flow change information of the light-field image, and then a high-resolution light field is reconstructed by using the offset of each unit image, so as to effectively improve spatial resolution.
Abstract translation: 本发明提供了一种用于处理光场图像的方法和装置。 该方法包括:获取光场图像; 获取所述光场图像中的每个单位图像的遮挡标签功能; 根据所述光场图像中的每个单位图像的强度变化信息和每个单位图像的遮挡标签功能的梯度变化信息,确定每个单位图像的偏移估计的成本函数; 通过使用成本函数获取光场图像中的每个单位图像的偏移量; 并通过使用每个单位图像的偏移来重构高分辨率光场。 首先可以根据光场图像的光流变化信息获取光场图像中的每个单位图像的偏移,然后通过使用每个单位图像的偏移来重建高分辨率的光场,从而 有效提高空间分辨率。