Local ternary pattern alphabet shape features for stone texture classification

Abstract

Representation of shape or its description is one of the challenging problems in computer vision and processing image. This is required in many application areas such as image classification, its analysis, interpretation, etc. The texture is one of the important features which can be used to identify regions or objects of consideration in an image. The proposed paper derives a set of features using textures based on Local Ternary Pattern (LTP) and shape components extracted from the basic texture elements of a 3 × 3 mask. The present paper utilizes six alphabet shape components on stone textures. Based on the frequent occurrences of shape components on stone textures, the proposed method classifies the texture into four classes, i.e., Marble, Mosaic, Brick, and Granite. The present paper is experimented on Paul Bourke, Mayang, Google, and VisTex stone texture databases consisting of four texture classes and achieved good classification results.