Two New Scale-Adapted Texture Descriptors for Image Segmentation
In texture segmentation it is key to develop descriptors which provide acceptable results without a significant increment of their temporal complexity. In this contribution, we propose two probabilistic texture descriptors: polarity and texture contrast. These descriptors are related to the entropy of both the local distributions of gradient orientation and magnitude. As such descriptors are scale-dependent, we propose a simple method for selecting the optimal scale. Using the features at their optimal scale, we test the performance of these measures with an adaptive version of the ACM clustering method, in which adaptation relies on the Kolmogorov-Smirnov test. Our results with only these two descriptors are very promising.
KeywordsImage Segmentation Optimal Scale Wavelet Frame Texture Segmentation Gradient Orientation
- 1.Carson, C., Belongie, S., Greenspan, H., Malik, J.: Blobworld: Image segmentation using Expectation-Maximization and its application to image querying. IEEE Trans. on Pattern Analysis and Machine Intelligence (2002)Google Scholar
- 3.Hofmann, T., Puzicha, J.: Statistical Models for Co-occurrence Data. MIT AIMemo 1625 Cambridge, MA (1998)Google Scholar
- 7.Knutsson, H., Granlund, G.: Texture analysis using two-dimensional quadrature filters. In: IEEE Computer Society Workshop on Computer Architecture for Pattern Analysis and Image Database Management, pp. 206–213 (1983)Google Scholar