ArtHistorian: an integrated indexing and personalized browsing system for art paintings

Bibliographic reference:

Multimedia Signal Processing and Pattern Recognition Lab, ARTHISTORIAN: AN INTEGRATED INDEXING AND PERSONALIZED BROWSING SYSTEM FOR ART PAINTINGS.

Available at: http://www3.itu.edu.tr/~sariel/publications/Cbmi_2005_1033.pdf

Extract:

3. CONTENT-BASED INDEXING

The design of an automatic painting classifier requires the mathematical modeling of the visual clarities as determining features. In this paper, the features that distinguish the five fundamental art movements, namely, classicism, impressionism, cubism, expressionism and surrealism, are defined. However, the same feature set can be effectively used for more than five classes.
In [5], it has been shown that the content of paintings can be indexed by exploiting six different features. Suppose μi refers to ith individual entity of the feature vector μ for i=1,2,.,6. Physically, μ1 is the percentage of dark colors. On the other hand, μ2 is calculated from the gradient map of the painting image, and referred as “gradient coefficient”. Especially, the “classicism” and “cubism” movements, and thereby, “classicist” and “cubist” painters can be classified quite accurately by using these two features. Classification performances of surrealist and expressionist paintings are augmented by μ2.
The classification performances of paintings that belong to impressionism and classicism movements are augmented by the inclusion of two extra features: μ3 is the number of local and global maxima in the luminance histogram. Moreover, μ4 specifies the color range that corresponds to the peak point of the luminance histogram. Impressionist paintings generate one maxima in the bright regions of the luminance histogram, whereas classicist paintings generate a maximum point in dark regions. In cubist, expressionist and surrealist paintings, more than one maximum point occur in variable regions of the histogram.

Dublin Core Metadata:

Title : ARTHISTORIAN: AN INTEGRATED INDEXING AND PERSONALIZED BROWSING SYSTEM FOR ART PAINTINGS.
Creator : Oguz Icoglu, Bilge Gunsel, and Sanem Sariel
Subject : ArtHistorian/ indexing system/ art paintings/ classification
Description : "This paper introduces ArtHistorian, an art painting indexing system designed for automatic generation of dynamic painting presentations. The developed system aims personalized content browsing with a classification based indexing and query method implemented from the art historians’ perspective. ArtHistorian represents the visual content of paintings by a 6-D feature vector that is robust to scale changes and can handle variations in lighting conditions. A nonlinear SVM classifier included in the system learns the characteristics of fundamental art movements, painters and painting styles. Automatically generating indexes based on the classification results, the system creates an efficient storage environment that enables the user to run content-based queries and to retrieve materials from extensive painting records. Thus, it is very suitable for Internet art collections and active Web museums."
Publisher : Multimedia Signal Processing and Pattern Recognition Lab
Date : 2005
Type : Text
Format : pdf
Identifier : http://www3.itu.edu.tr/~sariel/publications/Cbmi_2005_1033.pdf
Source: None
Language : En
Relation : -
Coverage : World
Rights : Electrical-Electronics Eng. Faculty. Istanbul Technical University