TEXTURE DESCRIBTION METHOD IN FREQUENCY DOMAIN, AND TEXTURE-BASED RETRIEVAL METHOD

摘要

PROBLEM TO BE SOLVED: To more accurately describe the texture of a video by describing the texture descriptor of the video through the use of a feature value extracted from the respective frequency domains of frequency domain division layouts in the form of a feature vector. SOLUTION: When an arbitrary video is inputted, the inputted video is Randon-transformed. The obtained set of primary Randon transform functions is Fourier-transformed. When a video function is Randon-transformed and is Fourier-transformed, it is moved to a pole coordinate frequency domain and a video signal is transformed into a pole coordinate frequency domain. The texture feature of the video is extracted from the pole coordinate frequency domain having a prescribed frequency sampling structure. A pole coordinate system frequency domain division layout is used. In the pole coordinate system frequency domain division layout, a frequency domain is divided by a human sight recognition system. The texture descriptor of the video, which describes the texture of the video, namely, a feature vector is calculated form the feature value.