Web inspection using multi-scale wavelet decomposition

        A wavelet based texture characterization and defect detection system has application in robust on-line inspection system. The objective of using multiscale wavelet representation is to attenuate the background texture and accentuate the defects. The Daubechies D2 filters were selected because it is efficient to implement them, and more importantly the structure of its corresponding 2-D kernels matches the plain weave pattern of the fabrics to be tested. Some of the results from the proposed algorithms are can be seen below.
  

Figure 1: Fabric samples with oil-stain and mispick and their corresponding segmented defects in using multiscale wavelet decomposition.

                   

 Figure 2 : Some more fabric samples with defects and their corresponding segmented defects.

Back to research page