Analyzing Space-Variant Textures Using Gabor Wavelets

概要

論文の詳細を見る
The problem adressed in this paper is the segmentation of textured regions with spatially varying textures. Those textures can be characterized by localized dominant frequencies, described here as emergent frequencies. The emergent frequencies vary smoothly, however this variation could be considerable over the entire image domain. Examples of such textures include textures projected from curved or deforming surfaces, flow-like and evolutionary patterns. The wood grain is another representative texture of this kind. Gabor wavelets are used to derive a constraint equation. The estimation of emergent image frequencies leads to an ill-posed problem which can be regularized using smoothing constraints. However, the smoothness assumption is violated at the boundary between texture regions. To detect these discontinuities, controlled-continuity stabilyzers are used as regularizing terms.
社団法人電子情報通信学会の論文
1997-08-13