ABSTRACT

For achieving better classification results in texture analysis, it is to combine different classification methods. Though there are existing methods which have been using fixed window size that resulted lack of classification accuracy and in order to improve the classification accuracy, the window size must be increased. Moreover the optimal window size is to be selected is also an important thing in the improvement of better classification output. In addition, since some classification techniques are used for micro textured structures and some are for large scale textured images, it is better to integrate different classification methods to achieve higher classification rate. This paper presents a new classification technique named unsupervised hybrid classification for texture analysis (UHCTA) that extracts the properties of different methods for achieving higher classification rate. Also comparison with the existing methods conform the merits of the proposed unsupervised hybrid classification for texture analysis method in terms of accuracy in various image conditions.