A Model-Based Approach to Semantic-Based Retrieval of Visual Information

Abstract

Visual context descriptor (VCD) is a new image representation scheme for visual content classification. It consists of a multidimensional vector in which each element represents the frequency of a unique visual property of an image or a region thereof. VCD utilizes the predetermined quality dimensions, such as types of features and quantization level, along with predetermined semantic model templates. The observed visual cues and the contextually relevant visual features are proportionally incorporated in VCD. Contextual relevance of a visual cue to a semantic class is determined by using correlation analysis of ground truth samples. Such co-occurrence analysis of visual cues requires transformation of a real-valued visual feature vector, say a color histogram or a Gabor texture, into a discrete event, e. g., terms in the text domain.