Optimal Spatial Predicate Determination of a Local Binary Pattern Operator

R.A. Lizarraga-Morales, R.E. Sanchez-Yanez, and V. Ayala-Ramirez (Mexico)


Texel size, Local binary pattern, Image texture analysis, Image classification, Scale Invariant.


A number of applications of texture analysis involve the processing of texture images where the scale factor with respect to a reference image is unknown. In such cases, most approaches present limitations in their performance. In order to overcome problems arising from this situation, we need to develop scale invariant operators that could be adaptive with respect to the texel extent of the texture present in the image under test. Such operators will be a key element of a scale-invariant texture classification system. We propose to include a texel size detection step in order to determine the spatial predicate of a Local Binary Pattern (LBP) operator best suited for the analysis of the texture under test. If we apply the custom-sized LBP operator with the same number of samples to images containing the same texture but at different scales, LBP histograms show to be very similar. This fact is then exploited to implement a scale-invariant texture classifier. We present our approach by comparing its performance against other texture approaches using the same texture data set.

Important Links:

Go Back