Reconstructing occluded regions using fast weighted PCA

Reconstructing occluded regions of the object is to automatically detect the occluded regions and background in the image and reconstruct these regions using image interpolation. This paper proposes a novel occluded region reconstruction method using Fast Weighted Principal Component Analysis (FW-PCA). The computation time of the weighted PCA can be reduced by using only the effective regions when calculating the principal component scores. The occluded regions are accurately detected by recursively updating the weight for each pixel in the image using FW-PCA. Then, the occluded regions can be reconstructed using the final weight. Thorough a set of experiments, we demonstrate that the proposed method exhibits higher performance than the conventional method.