Title: An approach for multiscale analysis of visual texture
Abstract: The main goal of multiscale analysis of visual texture is to extract the same texture descriptors from two images at different scales, but most of the methodologies obtain different texture descriptors. The effect of support area selection on the Local Binary Pattern (LBP) operators is discussed here for scale invariant texture descriptor extraction. In this work, determination of the fundamental pattern (texel) is required as a scale reference in the image and it is used as a pre-analysis to select the size of a suitable operator. A new non-parametrical operator, based on the LBP operator, and spiral-shaped adaptive operator is also proposed for multiscale texture analysis. Such a shape allows the operator to cover a larger area on the field of view using a predefined number of samples. In order to test the performance of our approach, feature vectors for sets of both synthetic and natural texture images at different scales are compared. Similarity is measured with the Cosine-Amplitude test. Feature vectors extracted with the operator proposed can be used as a texture descriptor in different applications of visual analysis like classification or segmentation tasks.
Publication Year: 2009
Publication Date: 2009-11-01
Language: en
Type: article
Indexed In: ['crossref']
Access and Citation
AI Researcher Chatbot
Get quick answers to your questions about the article from our AI researcher chatbot