System of textural features based on the measurement of spatial frequencies

Abstract

Listen

Translate article

The paper indicates that one of the forms of contextual information is the concept of texture, which is a functional of a set of fragment points. It is noted that the advantage of textural features is the potential for the aggregation of contextual information with certain properties of invariance to a specific target of patterns recognition. The informative significance of textural features as compared to the spectral features is substantiated and an overview of texture features is given. In this regard, the problem of features selection is analyzed and the distribution of textural features into groups is shown. A group of features is studied that are based on measurements of the spatial frequencies and the feasibility of applying groups of features based on the use of the continuous Fourier transform and its discrete analogue is shown. It was found that the combined application of features based on the use of the continuous Fourier transform and its discrete analog allows making the features system sensitive both to the dimensions and to the orientation of elements forming the texture. The conclusion was made on the feasibility of application of the features, based on measurements of the spatial frequencies in various fields of human activities associated with patterns recognition on the images.