Abstract: In order to improve the defect detection efficiency on magnetic tile surface, prevent defect undetected, an automatic detection method is proposed for the problem that defects with dark color and low contrast on magnetic tile surface can not be extracted via the classical defect detection algorithm. This method is based on adaptive threshold surface in nonsubsampled Contourlet domain.According to the characteristics of NSCT sub-band coefficients in different regions, the threshold of the NSCT coefficients is changed by using line-scan model, and then the normalized adaptive threshold surfaces are formed for different scales and different directions.Compared with the original NSCT sub-band coefficients, a series of new NSCT sub-band coefficients are formed.Finally the new NSCT coefficients are reconstructed and the defects on magnetic tile surface are extracted.The experimental results indicate that the proposed method can remove texture on magnetic tile surface effectively and the accuracy rate of extraction defect achieves 95%.