Local Texture Patterns - A Univariate Texture Model for Classification of Images

Texture is an important spatial feature useful for identifying objects or regions of interest in an image. In this paper a new spatial method of texture model is proposed and a comparison of some of the histogram based texture models is carried out. This paper presents a texture analysis operator based on 'local texture patterns (LTP)'. The global image texture is modeled by 'pattern spectrum' which is an occurrence histogram of LTP over the image. The operator is designed as a gray-scale and rotational invariant texture measure on a local neighborhood. The efficiency of the proposed texture model is tested with classification of images based on log-likelihood similarity measure. The results show that the proposed method provides a very good and robust performance.