Texton is a extensively applied approach for texture analysis. This technique shows a strong dependence on certain number of parameters. Unfortunately, each variation of values of any parameter may affect the texture characterization performance. Moreover, micro structure texton is unable to extract texture features which also have a negative effect on the classification task. This paper, deals with a new descriptor which avoids the drawbacks mentioned above. To address the above, the present paper derives a new descriptor called Fuzzy Based Texton Binary Shape Matrix (FTBSM) for clear variation of any feature/parameter. The proposed FTBSM are defined based on similarity of neighboring edges on a 3×3 neighborhood. With micro-structures serving as a bridge for extracting shape features and it effectively integrates color, texture and shape component information as a whole for texture classification. The proposed FTBSM algorithm exhibits low dimensionality. The proposed FTBSM method is tested on Vistex and Akarmarble texture datasets of natural images. The results demonstrate that it is much more efficient and effective than representative feature descriptors, such as logical operators and GLCM and LBP, for texture classification.