Filtering for texture classification: a comparative study

Abstract

In this paper, we review most major filtering approaches to texture feature extraction and perform a comparative study. Filtering approaches included are Laws masks, ring/wedge filters, dyadic Gabor filter banks, wavelet transforms, wavelet packets and wavelet frames, quadrature mirror filters, discrete cosine transform, eigenfilters, optimized Gabor filters, linear predictors, and optimized finite impulse response filters. The features are computed as the local energy of the filter responses. The effect of the filtering is highlighted, keeping the local energy function and the classification algorithm identical for most approaches. For reference, comparisons with two classical nonfiltering approaches, co-occurrence (statistical) and autoregressive (model based) features, are given. We present a ranking of the tested approaches based on extensive experiments.

A Generalized Local Binary Pattern Operator for Multiresolution Gray Scale and Rotation Invariant Texture Classification

“…Gray scale invariance is also necessary if the gray scale properties of the training and testing data differ. This was clearly demonstrated in our recent study [9] on supervised texture segmentation with the same the image set that was used by Randen and Husoy in their recent extensive comparative study [12]. However, real world textures with a large tactile dimension can also exhibit non-monotonic intensity changes, e.g.…”

Texture Edge Detection Using the Compass Operator

“…However, there is no clear consensus on which statistical methods work best. Randen and Husoy undertook a quantitative evaluation of statistical and frequency-based approaches to texture segmentation [4]. Their conclusion was that none of the methods they tested worked well on a large variety of textures.…”

“…The distributions of these patterns should be consistent over an area of uniform texture, provided that the distribution is observed at a scale larger than the inherent texton size. Examples of oriented filters previously used in texture analysis include spatial filters such as Laws filters or Gabor filters and frequency-space filters such as Fourier, discrete cosine and wavelet transforms [4].…”

Experimentation on the Use of Chromaticity Features, Local Binary Pattern, and Discrete Cosine Transform in Colour Texture Analysis

“…For comparison, we selected the discrete cosine transform suggested by Ng et al [7]. Randen and Husoy [11] found relatively good results using the DCT approach. Local linear properties can be extracted using well known transforms similar to DCT.…”