关键词: 深度估计, 散焦模糊量, 超像素分割

Abstract:

Existing depth recovery algorithms for single defocused image may be complex and not effective at edge locations, textured areas and the regions with soft shadows. This paper proposes a depth estimation from single defocused image based on superpixel segmentation. At first, original image is segmented to some superpixel modules, and then the blur of these modules can be computed based on the blur of pixels at edge locations. In so doing, sparse depth maps at superpixel level will be obtained and optimized. At last, truthful and accurate depth maps may be generated. The proposed method can decrease the error to the minimum and simplify the extension process of blurs from edges to all image pixels. The experimental results on real data demonstrate that the proposed method is not only faster, but can improve depth maps effectively, especially for edge locations, textured regions and soft shadows as well.

Key words: depth estimation, defocus blur, superpixel segmentation