Estimation of Degree of Attractiveness of Each Part of Images based on Principal Component Analysis for Images taken by Plural Cameras

This paper describes a method that estimates the degree of attractiveness of each part of input images taken by plural cameras. The attractiveness is expected to indicate the high possibility of existing of objects. We assume that the objects are included in only a few images among plural images. The degree of attractiveness of each part of the particular input images including objects is obtained by subtracting the reconstructed background images from the input images. The adequate background images for the particular input images are inferred from other input images which dose not include any objects. Since illumination condition is inferred from the input images which dose not include objects, appropriate background images can be reconstructed without suffering the effect of included objects. The effectiveness of the method is confirmed by computer simulations.