Disparity estimation based on integral imaging in sub-pixel resolutionusing maximum a priori (MAP) registration

摘要

A sub-pixel resolution disparity estimation method using MAP registration in elemental image based on integral imaging is proposed. Accurate depth map extraction method is an important research issue in the latest researches about three-dimensional (3D) content developments. We capture a 3D object using lens array, and generate elemental images. Many previous researches calculated disparity of each elemental image using sum of square distance; however, the accuracy of disparity is limited by one pixel unit. For enhancing accuracy of estimated disparity, we adapt the maximum a priori (MAP) registration algorithm to elemental image set. The proposed method can calculate each disparity of elemental image set in sub-pixel resolution using MAP registration optimization.