Improving the discrimination capability with an adaptive synthetic discriminant function filter

Abstract

In this paper a new adaptive correlation filter based on synthetic discriminant functions (SDF) for reliable pattern recognition is proposed. The information about an object to be recognized and false objects as well as background to be rejected is used in an iterative procedure to design the adaptive correlation filter with a given discrimination capability. Computer simulation results obtained with the proposed filter in test scenes are compared with those of various correlation filters in terms of discrimination capability. © Springer-Verlag Berlin Heidelberg 2005.