Analysis on preprocessing methods for typical defect images in weld X-ray inspection

To enhance the accuracy of image recognition for typical defects in weld X-ray inspection, it is necessary to employ effective image preprocessing methods to improve the image quality of radiographic films in weld X-ray detection. By integrating the characteristics of weld defects, a variety of filtering and image enhancement preprocessing algorithms were investigated. Moreover, a comprehensive evaluation of the preprocessing effects was carried out using a combined subjective and objective evaluation system. Experimental results showed that the parallel filtering algorithm based on adaptive median filtering and wavelet transform exhibited the best overall performance in noise reduction. Meanwhile, the Contrast Limited Adaptive Histogram Equalization algorithm showed significant advantages in image enhancement, effectively improving the contrast and discernibility in defect regions.