Texture Classification using Angular and Radial Bins in Transformed Domain

Arun Kulkarni,Bishrut Bhattarai,Ashim Sedhain,Aavash Sthapit,Saurav Panthee

doi:10.14569/ijacsa.2021.0120301

Abstract

Texture is generally recognized as fundamental to perceptions. There is no precise definition or characterization available in practice. Texture recognition has many applications in areas such as medical image analysis, remote sensing, and robotic vision. Various approaches such as statistical, structural, and spectral have been suggested in the literature. In this paper we propose a method for texture feature extraction. We transform the image into a two-dimensional Discrete Cosine Transform (DCT) and extract features using the ring and wedge bins in the DCT plane. These features are based on texture properties such as coarseness, smoothness, graininess, and directivity of the texture pattern in the image. We develop a model to classify texture images using extracted features. We use three classifiers: the Decision Tree, Support Vector Machine (SVM), and Logarithmic Regression (LR). To test our approach, we use Brodatz texture image data set consisting of 111 images of different texture patterns. Classification results such as accuracy and F-score obtained from the three classifiers are presented in the paper.

Highlights

Texture is generally recognized as being fundamental to perception
We used Brodatz texture image data set consisting of 111 images of different texture patterns [24]
Most of the information in the image is concentrated in a few coefficients that are in the top left corner of the Discrete Cosine Transform (DCT) matrix

Summary

INTRODUCTION

Texture is generally recognized as being fundamental to perception. Texture provides useful information in identifying objects in images. Texture primitives may be pixels or aggregate of pixels such as regions. Texture descriptors provide measures of properties such as smoothness, coarseness, and regularity [3]. Gonzalez and Woods [4] describe three principal approaches for texture analysis: statistical, structural, and spectral. Statistical approaches yield texture properties such smoothness, coarseness, or graininess. We propose a new algorithm for extracting texture features from the two-dimensional Discrete Cosine Transform (DCT) of the image. These features capture directional and coarseness properties of the texture. We classify texture images using these features with statistical models.

RELATED WORK

PROPOSED APPROACH

EXPERIMENT AND RESULTS

CONCLUSIONS