A novel enhancement method of X-ray image based on multi-scale adaptive fusion

Abstract

To make X-ray image detail enhancement method have good performance, be easier to implement, be suitable for engineering applications, and be able to run on general computers without strong computing power, a novel detail enhancement method of X-ray image for non-destructive testing (NDT) is proposed based on the multi-scale adaptive fusion. The proposed method takes the variation of the Gaussian convolution X-ray image relative to the original X-ray image as the core, employs the sigmoid function and tanh function to compute the pixel importance and the nonlinear correlations across different gray levels respectively, and complementarily fuses their calculation results as the data fusion coefficients for the multi-scale adaptive enhancement. The experimental results show that the proposed method effectively improves the peak signal to noise ratio (PSNR), structural similarity index measure (SSIM) and clarity (CL) of the X-ray image compared to the methods based on histogram equalization (HE) and contrast limited adaptive histogram equalization (CLAHE). These results underscore the efficacy of multi-scale adaptive fusion for enhancing the X-ray image, suggesting a promising direction for future research in this field.