An Algorithm for Texture Removal and Defect Extraction

It is challenging to inspect a metal part with textures because the conventional defect detection methods cannot effectively distinguish between texture features and defect features, which can easily lead to misjudgment of defects. In this paper, we propose a new method for detecting surface defects which can effectively eliminate the texture features' impact on defect detection. By computing the average energy of directional length and average gray scale energy, the distribution of the energy difference in the whole image is obtained. According to the distribution of the energy difference, textures' impact on defects detection can he effectively removed and the surface defects can then be extracted. Experiments were performed to validate the proposed method and the results demonstrate the proposed method can effectively remove surface textures and accurately detect surface defects.