Western Painting Faction Classification with Multiple Image Characteristics

• Main Authors: CHEN, YOW-SHIN, 陳又新
• Other Authors: Wang, Chung-Shing
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/45427133799474106299

碩士 === 東海大學 === 工業設計學系 === 102 === Impressionism, cubism and futurism are the three kinds of factions which are very important in the modern art history. Enormous painting artworks had created from these classes, but the research about auto-classifying the kind of faction the painting is belonged to is rare. This paper studies automatic classification on the three kinds of classical western paintings. An effective method for automatic image classification has been proposed. A set of multiple characteristics of the image collections is integrated to support the method. This paper has accomplished the followings: 1. To collect, study and analysis enough number of digitized paintings out of the three different factions the images, and propose the multiple image characteristics set. 2. To apply image processing techniques such as edge detection, image entropy, Hough transform and corner detection etc. to establish the characteristics set. 3. To establish a neural network model as the classifier, train the network with the characteristics set, and create the classifier system. 4. To apply the classifier to new images to verify the system’s performance. 5. To reach a satisfactory level of the accuracies for the classifier, i.e., 92% for impressionism classification, 85.71% for cubism classification and 64.52% for futurism classification