Evaluation of Visual Features of Textile Designs Using Genetic Algorithm

Abstract

Designs that have been stored in early days are often used as a reference when textile designs are created. To effectively use these resources, an well-organized design database must be constructed and a useful searching method must be studied. As to design searching method, searching methods that deal with human subjective information is requested. As a technique for these searching methods, one method to characterize visual feature of textile designs is proposed. In this study, a correspondence was examined between the subjectivity like "strong impact" for textile designs and spatial frequencies that are objective visual features for them. In this examination, genetic algorithm was introduced in order to increase the efficiency of processing time. It was presumed that the subjective impressions of "strong impact" were related to lower components of spatial frequencies of textile design images. In this result, the textile designs that impress human when human compare designs are similar to the textile designs that are made up of lower spatial frequency components. As to design searches from subjective impressions of "strong impact," lower spatial frequency components that are objective visual features are available for one of subjective searching item.