Traditional ethnic costume pattern segmentation method based on feature matching algorithm

Guodong Zhang
Hubei Institute of Fine Arts, Wuhan, 430205，China

Abstract:

This paper presents a novel traditional ethnic costume pattern segmentation method based on feature matching algorithm, which combines scanning and recognition techniques. For the traditional ethnic costume, the process of scanning is divided into two parts: from photographing to image processing, from image processing to segmentation. The first step starts with photographing the total picture of a garment through optical system in order to acquire photo color picture; then, it processes photo color picture for appropriate size and arranges in correct order by lines or stitches; finally it converts photo color picture into binary black-and-white photograph by using edge detection operator. In the second step, it extracts attribute features from segmented image then computes feature matching key points, the key points classified into several groups. Then it establishes a new parameter which is called region similarity that is based on distance between key points. Finally it detects the actual cutting region of traditional ethnic costume pattern quickly and accurately by using new parameter.