Automatic depth map generation from a single image using segment-adaptive depth merging

Sung In Cho, Kyoung-Rok Cho, Young Hwan Kim

Abstract: This paper proposes an advanced approach to learning-based depth map generation for automatic 2D-to-3D conversion, which utilizes 3D histogram-based segmentation for a segment-adaptive depth merging. By using a segment-adaptive depth merging, the proposed method successfully enhances the accuracy of the conventional learning-based depth map generation.