Performance-Enhanced Three-Dimensional Object Recognition by Computational Integral Imaging with Depth Data of Picked-Up Elemental Images

A highly performance-enhanced three-dimensional (3D) object recognition system has been implemented using a computational integral imaging technique with depth of picked-up elemental images. For this, depth data of the picked-up elemental images are extracted, and elemental images are classified into a number of groups depending on these depth values. Then, each group of depth-dependently classified elemental images is separately used for reconstructing the object image on each corresponding depth plane. By this method, defocused images resulting from the objects located on other depth planes, which unavoidably occur in the conventional computational integral imaging reconstruction (CIIR) method, can be effectively removed. That is, in the proposed method, only the clearly focused object images can be reconstructed where the objects were originally located. Thus, a significant performance improvement of the CIIR-based 3D object recognition system is obtained. Experimental results confirm the feasibility of the proposed method.