The crack defect area of magnetic tile is small, the contrast is low, there is cutting texture interference in the background, and the brightness distribution is uneven. A detection method based on contour wave and matrix singular value gradient difference (CT-SVD-DG) is proposed to address the challenges of traditional visual segmentation methods in extracting image features, low defect localization and segmentation accuracy, high misjudgment rate, and poor real-time performance. This algorithm first performs high-order contour wave transformation on the magnetic tile image to obtain sub-band coefficients of different scales and directions; Perform singular value decomposition on the sub-band coefficient matrix to obtain a singular value sequence; To accurately obtain the filtering threshold for singular values, the article proposes the concept of singular value gradient difference, which is beneficial for distinguishing defects and backgrounds more accurately by gradient difference; Finally, the enhanced image is obtained using inverse contour wave transformation. The experimental results show that this method can effectively filter out magnetic tile background texture, noise, and interference while protecting crack features. Compared with similar algorithms, the method proposed in this paper has significant advantages in detecting surface small crack defects. The average values of the Se and Dice indicators reach 94.3% and 90.4% respectively, while the Ac indicator reaches 99.6%. The detection efficiency is doubled, and it has strong engineering practical value.