Fast Matching Algorithm between Line Pictures Described by Chain Codes Applied for Human Face Identification

Abstract

A new matching algorithm for line pictures described by chain coded expression is herein presented. In personal identification system, which uses faces, three-dimensional facial features are described by isodensity lines extracted from a face. The identification is then executed by composing the similarity between two isodensity lines. In the conventional system, the matching procedure is executed based on an image based matching algorithm. From a practical viewpoint, however, the matching time must be reduced as much as possible, while maintaining higher discriminatory accuracy. Because isodensity lines are the line pictures, these line pictures can be transformed into chain codes. By using a chain coded expression of the isodensity lines, remarkable improvement in the processing performance can be expected in terms of matching time and memory requirements. In the conventional system, many techniques such as Pyramid Structure of an image, SSDA (sequential similarity detection algorithm) and a multistage matching algorithm were introduced to reduced matching time. In the chain coded expression of the isodensity lines, however, if we try to adopt these time reducing techniques used in the conventional system, all matching processes must be executed using only chain coded expressions. In general, however, it is quite difficult to perform. Therefore, new matching techniques, which are adapted for chain coded expressions, must be devised. In this paper, a new matching algorithm for chain coded line pictures is presented, as well as the construction method of Pyramid structure for line pictures. From the experimental results, it was proven that both the matching time and memory requirements could be reduced by approximately one twentieth and one ninth, respectively