Using Polygons to Recognize and Locate Partially Occluded Objects
IEEE Transactions on Pattern Analysis and Machine Intelligence
| Hi-index | 0.00 |
Computational integral imaging is a promising technique in partially occluded 3D object recognition. With elemental images (EIs) of partially occluded 3D object, the plane image of 3D object to be interested was reconstructed at the location where 3D object was originally located using a computational integral imaging reconstruction (CIIR) algorithm. However, occlusion prevents the high-resolution reconstructed image due to superimposing its defocusing and blurred image at the same time. To overcome this problem, in this paper, we propose a novel occlusion removal method of partially occluded 3D object in computational integral imaging. In the proposed method, we use the variance of ray intensity distribution emitting from EIs and then a series of variance plane images from the EIs is used to estimate the area and distance of occlusion since the intensity variance of focused plane image for occlusion is the lowest at the occlusion location. On the basis of the extracted information, occlusion in the EIs is simply eliminated. Then, the plane images are reconstructed with the CIIR algorithm for the occlusion removed EIs, in which we can obtain the improved high-resolution plane image. To show the feasibility of our proposed scheme, some experiments were carried out and its results are presented as well.