2011 | OriginalPaper | Buchkapitel
Volume Local Phase Quantization for Blur-Insensitive Dynamic Texture Classification
verfasst von : Juhani Päivärinta, Esa Rahtu, Janne Heikkilä
Erschienen in: Image Analysis
Verlag: Springer Berlin Heidelberg
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
In this paper, we propose a blur-insensitive descriptor for dynamic textures. The Volume Local Phase Quantization (VLPQ) method introduced is based on binary encoding of the phase information of the local Fourier transform at low frequency points and is an extension to the LPQ operator used for spatial texture analysis. The local Fourier transform is computed efficiently using 1-D convolutions for each dimension in a 3-D volume. The data achieved is compressed to a smaller dimension before a scalar quantization procedure. Finally, a histogram of all binary codewords from dynamic texture is formed. The performance of VLPQ was evaluated both in the case of sharp dynamic textures and spatially blurred dynamic textures. Experiments on a dynamic texture database DynTex++ show that the new method tolerates more spatial blurring than LBP-TOP, which is a state-of-the-art descriptor, and its variant LPQ-TOP.