Image screening based on projection pursuit for statistical image recognition

Abstract

AbstractA preprocessing method, called image screening, is presented to improve the recognition rate and efficiency in statistical image recognition. The problem of detecting a specified object in input images is treated as a two‐class classification problem in which the image falls into both a set of subimages in the target object (figure) class and the other set of subimages in the ground class. An image screening algorithm based on projection pursuit selects a candidate set of subimages that is similar to the object class, it rejects the remaining set using screening filters whose design is based on projection pursuit. The feature space for recognition is obtained from the selected subimages. Two kinds of measures to evaluate the performance of image screening are defined. The error rate in image screening is related to the total recognition rate of the system and the rejection rate of the noise image is related to the recognition efficiency. Two kinds of experiments were conducted, one to detect the eye and mouth areas in a face image and the other to detect the text area in a document image. Experimental results for these two tasks demonstrate that our method improves the recognition accuracy and efficiency.