Abstract: Aiming at the problem that  the detail scales in X-ray images were diverse, it was difficult to meet the enhancement requirements of different scale details simultaneously using a single enhancement strategy, we proposed  a multi-level fuzzy enhancement method for X-ray images based on wavelet phase filtering. Firstly, by using histogram equalization to redistribute grayscale values to improve contrast, and utilizing linear interpolation to adjust the grayscale values of blurry local areas based on clear pixel information around them to restore details. Secondly, after optimizing the grayscale distribution, the noise points in the X-ray image were more significant, we first used  the Mallat algorithm to perform two-dimensional wavelet decomposition on the adjusted image, and then combined  wavelet phase filtering to achieve image denoising, enhance image clarity, and optimize image quality. Thirdly, we converted the clarity enhanced X-ray image into a histogram format to query peak gray levels, defined fuzzy sets and membership functions, used fuzzy subsets as enhancement levels, optimized membership using power functions and multi-level enhancement coefficients, and then traversed gray levels to complete membership optimization, and achieved image detail enhancement. Finally, we reverted the  fuzzy domain changes back to the grayscale domain to obtain grayscale update results, achieving multi-level fuzzy enhancement, highlighting image details and lesion areas, and improving diagnostic accuracy. The experimental results show that the proposed method can effectively highlight the detailed information in X-ray images. At different grayscale levels, the frequency of X-ray images remains between 18 Hz and 35 Hz, with relatively smooth frequency fluctuations. The mean gradient (MG) value can be maintained above 0.35. The proposed method has a significant enhancement effect on X-ray images, which helps to accurately and effectively analyze the lesion area, thereby improving the accuracy and efficiency of disease diagnosis.

Key words: histogram equalization algorithm, linear interpolation algorithm, Mallat algorithm, two-dimensional wavelet decomposition, membership function