Discrete Wavelet Frame Representations of Color Texture Features for Image Query

T Chen,KK Ma,LH Chen
DOI: https://doi.org/10.1109/mmsp.1998.738911
1998-01-01
Abstract:We propose a wavelet-based multi-channel scheme to extract human-perception relevant color texture features for image indexing and querying. For each spectral band, a two-dimensional discrete wavelet frame (DWF) decomposition is applied first, followed by an enveloping operation performed on the resulting wavelet coefficients. The unichrome features computed from the enveloped coefficients of the individual band as well as the opponent features that provide the spatial correlation between different spectral bands are jointly exploited for accurate image classification. The experimental results are promisingly, showing that the proposed approach is suitable for browsing color texture images
What problem does this paper attempt to address?