Enhanced Dynamic Medical Images In Real Time

As medical image is widely used in clinical diagnosis and has a huge development potential, dynamic real-time medical image enhancement method has been widespread concerned. However, there are still two fundamental problems in dynamic real-time medical image enhancement:1. The contradiction between contrast,detail enhancement and noise reduction which blurs the image detail; 2, the contradiction between computational algorithms and requirement of real-time processing. In order to solve these contradictions, a number of methods have been proposed for medical images based on the traditional image processing techniques by many researchers, also some algorithm has made some kind of simplified to meet the real-time processing requirements, and achieved good results in some extend, but there are still much shortage in the enhance effect.In this paper, we propose a new denoising algorithm which is based on the noise level classification and reliable motion estimation, and synthesize to achieve a better enhance result for dynamic real-time medical image. Compared with the recent methods based on the sparsity of image or on the denoising template, our denoising algorithm is based on noise level classification and reliable motion estimation. To be specific to say, we estimate the noise level of image with the image pyramid, then we estimate the motion parameters based on different pyramid level according to different noise level, average the weighted multi-frame dynamic image with the motion parameters to get a good denoising result.In the part of contrast and edge enhancement, we synthesize various methods in space-field and wavelet-field to meet various processing need for different part of the body. To be specific to say, we use the Contrast Limit Adaptive Histogram Equalization algorithm, wavelet enhancement algorithm based on Cohen Daubechies Feauveau wavelet family and the edge enhancement algorithm based on the Laplacian image pyramid to enhance the image. To solve the contradiction between computational algorithms and requirement of real-time processing, we use the Compute Unified Device Architecture to parallel these methods to meet the requirement.Our dynamic real-time medical image enhance system can process the dynamic image which is 1000*1000 pixels with 16 bit gray scale, attaining the speed about 33 frame per second. In practice, our enhance result receive much recognition by many doctors and the system has been used in the department of radiology in many hospitals.