Abstract

Texture is an attribute which is so easy to be visually recognized but hard to be mathematically characterized. In\nthis paper, a grating cell texture operator has been proposed that responds vigorously to texture present in the form of grating\nof bars of appropriate orientation, position and periodicity. The scene images similarly represent a consistent texture in\nnatural objects (e.g. mountain, grass, trees etc.) which respond to the grating cell. In the images, the texture of different\nobjects of scene varies from one another. Thus, the grating cell responds differently to different objects of a scene image for a\ngiven set of parameters which is the criteria for segmentation. The textural properties of an image are stored in a feature\nmatrix which is given to k-means clustering algorithm for segmentation purpose.