Defect Detection of Tiles with Combined Undecimated Wavelet Transform and GLCM Features
Afsane Fathi1, Amir Hassan Monadjemi2, Fariborz Mahmoudi3

1Afsane Fathi, Department of Computer Engineering, Islamic Azad University, Qazvin branch, Qazvin, Iran.
2Amir Hassan Monadjemi, Department of Computer Engineering, Isfahan University, Isfahan, Iran.
3Fariborz Mahmoudi, Department of Computer Engineering, Islamic Azad University, Qazvin branch, Qazvin, Iran.

Manuscript received on May 01, 2012. | Revised Manuscript received on May 04, 2012. | Manuscript published on July 05, 2012. | PP: 30-34 | Volume-2, Issue-2, May 2012. | Retrieval Number: B0499032212/2012©BEIESP
Open Access | Ethics and Policies | Cite 
© The Authors. Published By: Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (

Abstract: Development of an automatic defect detection system has a major impact on the overall performance of ceramic tile production industry. With this in mind, in this paper, a new algorithm has been offered for segmentation of defects in random texture tiles. firstly, by using undecimated discrete wavelet Transform (UDWT), frequency features of textures which are robust towards transition could be extracted. Then a co-occurrence matrices of sub-bands, in order to extract texture information, is obtained. Finally, after obtaining special characteristics from the combination of the two new methods, a back propagation neural network is applied for segmentation which is the final product of this. The results, both visually and computationally, show a higher accuracy while using this method than the conventional wavelet method and co-occurrence matrices that was utilized previously. The reason could be its independent from scale and rotation nature compared to the typical transform. Different locations of defects make different wavelet coefficients and ultimately increase the defect segmentation performance of a wide variety of defects.

Keywords: Defect detection, Wavelet Transform, Undecimated Wavelet Transform, Co-occurrence Matrices, Back-Propagation Neural Network.