Abstract
In this paper, we propose a partial occlusion removal method to obtain an artifact-free three-dimensional reconstructed image by computational integral imaging reconstruction (CIIR). The method utilizes the mutual constraints that exist between the elemental images in the elemental image array to obtain a reliable depth map. Normally, in CIIR, the elemental images are very small and have very low resolution, which makes it difficult to obtain a reliable depth field. Therefore, previous researchers proposed a subimage-based framework, where elemental images are first transformed into subimages by a subimage transform to increase the small sizes of the elemental images. After that, a stereomatching algorithm is applied on the subimages to compute the depth map from which the region of occlusion can be extracted. However, due to the inherent incorrectness of the subimages and the weak constraint used in the stereomatching algorithm, the resulting depth map includes many false regions, which result in a noisy segmentation of the occluding object region. By utilizing the mutual relationship between neighboring elemental images, a strong constraint can be applied on the computation of the depth map. This reduces the uncertainty in the computation and thus results in a reliable depth map. This is due to the fact that owing to the strong constraint, the depth map can be computed directly from the elemental images, excluding the incorrectness of the subimages. We also describe both the pixel-wise and window-based implementations of the proposed method. Experimental results verify the performance of the proposed method.
© 2015 Optical Society of America
Full Article | PDF ArticleMore Like This
Joon-Jae Lee, Byung-Gook Lee, and Hoon Yoo
Appl. Opt. 50(13) 1889-1893 (2011)
Miao Zhang, Yongri Piao, and Eun-Soo Kim
Appl. Opt. 49(14) 2571-2580 (2010)
Dong-Hak Shin, Byung-Gook Lee, and Joon-Jae Lee
Opt. Express 16(21) 16294-16304 (2008)