Research On The Methods Of Wavelet Transforms For DR Image Enhancement

With the development of computer technology and X-ray photography. Direct digitalX-ray photography system, because of its simple, effective, easy-to-image post-processingand other advantages, became one of the common inspection method in medical diagnosis.But, in the processing of DR system imaging or digitization. Because of the electronicdevices noise, optical scattering and photon scattering effects. Pathological diagnosis of thedoctor’s work has some difficulties. Therefore, the need for image enhancement to improveimage resolution and image edges details in the post-processing stage of DR system.In this paper, the main study contends of the DR system are the enhancementoperations in the image post-processing. The purpose is to enhance the image detailinformation, improve resolution, rich image contents. In the image enhancement, the majorproblem is noise effect. The traditional approach in the processing of image enhancementoften amplified noise together. So, the key point is how to denoising in the enhancementprocessing. This paper use wavelet transforms, decompose the image into different scalesin different directions on the high and low frequency signal. Through the use of denoisingand enhancement to high-frequency signal, get the enhanced image with reconstruction ofwavelet. Then, use enhanced post-processing operation through the image edgeinformation，get the final enhanced result image.The main research contents are as follows:1. Based on the nonlinear and soft threshold gain function, design a threshold-basednon-linear gain function. Adjust the gain factor to change the shape of the gain function,suppress noise; enhance image information, so the image outline becomes clear.2. Improve and simplify the traditional VisuShrink threshold method. Not need toestimate the noise variance, simplifies the operation steps of threshold selection. Thismethod can accord the scale numerical the localized threshold, also meet the noise energyin the wavelet domain decreases with the scale as the transformation law.3. After the wavelet transform use enhanced post-processing operation, with the resultsof edge detection, integrate the wavelet enhanced result and the unsharp masking result toget the final enhanced image. This operation increased enhancement effect to some extent,and effectively suppress the wrong phenomenon caused by enhanced or denoising.4. Implemented enhanced post-processing operations in parallel computing on the CUDA platform. With each GPU thread compute the corresponding pixel of the results.The efficiency of parallel computing methods than the CPU serial method underconventional increases27.3times, the acceleration effect is obvious.