主講人：黃文良 研究員 (中研院研究員)
時間：102年6月5日(星期三13:30-15:30)
地點：三峽校區人文大樓文1F11教室
題目：Multi-scale Depth Estimation from Blurring and Debluring of a Single Image
大綱：
We propose a novel depth estimation method that can derive a depth-map from a single image of a conventional camera. Our approach is based on estimating the blurriness of a patch by the combination of the blurring and debluring process. We formulate the problem as a
constrained optimization where the relation between the out-of-focus blurriness and the depth of a point is the constraint. We analyze the approach in noiseless and noisy environments and show that blurriness estimation methods can derive the depth of a non-smooth patch, but not a smooth patch. Thus, the depths of non-smooth
patches are propagated to smooth patches. We provide a multi-scale framework to improve the robustness for our approach. Our results are compared to that of the Bae and Durand's method, which uses Elder and Zucker's method to estimate depths on edges, and the
estimated depths are propagated to other area of an image. The experimental results show that our method is close to the `ground-truth' depth (measured by the Kinect camera) than the other method, even in noisy environment.
敬邀參加~
