The Detection Of Local Abnormalities In Random Macro Textures

Various techniques have been suggested for the detection of abnormalities in regularly textured patterns. These consist of searching for strong variations in a scene of replicated texture primitives. However, higlighting irregularities in randomly structured textures such as ceramic, granite or marble, is more challenging. We present a group of techniques which are based in both the spatial and feature domain and all detect abnormal formulations across the surface of randomly textured images. The techniques are based particularly on clustering in colour space, primary morphological operations, and connected component labeling. We conclude by explaining the purpose of using different techniques of varying complexity for the detection of the same type of faults, and where the use of a particular technique would be more appropriate. 1 Introduction Texture is generally divided into two main types: micro and macro textures. Many methodologies have been developed to describe and model them ..