Enhanced Wavelet Based Approach for Defect Detection in Fabric Images

Fabric defect detection is one of the indispensible units in the manufacturing industry to maintain the quality of the end product. Wavelet transform is well suited for quality inspection application due to its multi-resolution representation and to extract fabric features. This paper presents the comparison of three wavelet based models. These models include Tree structured wavelet transform, wavelet transform with vector quantized principal component analysis and Gabor wavelet network. The wavelet based models are combined with golden image subtraction to identify the fabric defect. The energy and entropy features are extracted and thresholding is performed to produce the binary image. The performance of the models is evaluated to verify the detection rate based on the segmented results. It can be concluded that the Wavelet transform with vector quantized principal component analysis provides better detection result.