AN EFFECTIVE STONE IMAGE CLASSIFICATION USING SURFACE PATTERNS BASED ON REDUCED DIMENSION AND GREY LEVEL RANGE MODEL

Abstract

This Texture classification plays an important role in many areas such as remote sensing, construction filed, medical imaging. The present paper proposes an innovative technique that classifies stone texture into four categories i.e. Brick, Marble, Granite and Mosaic based on the Surface Patterns (SP). Surface Pattern of the stone image is represented using Decreased Dimension and Reduced Grey level Range Matrix (DDRGRM) model. The DDRGRM model reduces the dimension of the image from 5×5 into 2×2 and also reduces the gray level range by preserving the significant feature information. The proposed technique focuses on the Surface Pattern (SP) which appears on DDRGRM image i.e. the patterns formed by Bezier curve (BC), U, V and T patterns. Hence, this model is termed as Surface Patterns on DDRGRM (SP-DDRGRM). The SPs on DDRGRM model of the stone texture are measured on a 5×5 sub image, because the surface patterns can be better viewed and they are difficult to fit in to 3×3 window. Based on the surface patterns count, the present paper develops a user defined algorithm for efficient classification that classifies stone image into 4 groups i.e. bricks, marble, granite and mosaic. The experimental analysis gives a clear idea about the features extracted in stone images by using SP-DDRGRM model which is suitable for both categories of classification algorithms i.e. standard algorithms like k-NNC, ID3 and SVM etc., and user defined algorithms. The experimental results indicate that proposed technique suits well for both standard and user-defined algorithms.