A study on a content-based image retrieval technique for Chinese paintings

Abstract

Purpose The purpose of this study is to build a database of digital Chinese painting images and use the proposed technique to extract image and texture information, and search images similar to the query image based on colour histogram and texture features in the database. Thus, retrieving images by this image technique is expected to make the retrieval of Chinese painting images more precise and convenient for users. Design/methodology/approach In this study, a technique is proposed that considers spatial information of colours in addition to texture feature in image retrieval. This technique can be applied to retrieval of Chinese painting images. A database of 1,200 digital Chinese painting images in three categories was built, including landscape, flower and figure. The authors develop an image-retrieval technique that considers colour distribution, spatial information of colours and texture. Findings In this study, a database of 1,200 digital Chinese painting images in three categories was built, including landscape, flower and figure. An image-retrieval technique was developed that considers colour distribution, spatial information of colours and texture. Through adjustment of feature values, this technique is able to process both landscape and portrait images. This technique also addresses liubai (i.e. blank) and text problems in the images. The experimental results confirm high precision rate of the proposed retrieval technique. Originality/value In this paper, a novel Chinese painting image-retrieval technique is proposed. Existing image-retrieval techniques and the features of Chinese painting are used to retrieve Chinese painting images. The proposed technique can exclude less important image information in Chinese painting images for instance liubai and calligraphy while calculating the feature values in them. The experimental results confirm that the proposed technique delivers a retrieval precision rate as high as 92 per cent and does not require a considerable computing power for feature extraction. This technique can be applied to Web page image retrieval or to other mobile applications.