Local Features in APICAS:Analyzing of Added Value of the Descriptors Based on MPEG-7 Vector Quantization

Abstract

An approach for extracting higher-level visual features for art painting classification based on MPEG-7 descriptors was implemented in the system “Art Painting Image Colour Aesthetics and Semantics” (APICAS). The approach consists of the following steps: (1) tiling images into non-overlapping rectangles in order to capture more detailed local information; (2) the tiles of the images are clustered for each MPEG-7 descriptor; (3) vector quantization is used to assign a unique value to each tile, which corresponds to the number of the cluster where the tile belongs, in order to reduce the dimensionality of the data. The distribution of significance of the attributes, the importance of the underlying MPEG-7 descriptors as well as analysis of spatial granularity for class prediction in this domain are analyzed. KeywordsContent-Based Image Retrieval (CBIR); Multimedia Semantics; MPEG-7 Descriptors; Vector Quantization; Categorization