Texture features based on the use of the hough transform and income inequality metrics

Abstract

Texture analysis of digital images has potential applications to image segmentation and classification. The quality of texture features can significantly determine the outcome of these two important applications. For images that consist of textures defined as patterns of straight lines, the use of features extracted using the Gray Level Co-occurrence Matrix (GLCM) is a popular choice. For each line that has a particular slope, one has to define a different predicate so that the matrix can capture that particular part of the texture. On the other hand, the Hough transform is a popular technique that detects lines that appear at different angles. We proposed an innovative way to extract texture information from the Hough accumulator using four income inequality metrics for patterns consisting of lines at different angles. We showed that when compared to four common texture metrics extracted from the GLCM, these new features can offer better quality. We used a feature selection algorithm and a classification example to illustrate the results obtained using these new income inequality texture metrics.