Research Of Medical Image Enhancement And Registration Based On Wavelet Transform Technology

Wavelet transform has the ideal of space/frequency multiresolution analysis. Wavelets offer a frequency/time representation of data that allows us time(respectively, space) adaptive filtering, reconstruction and smoothing. It can focus on any details of the analysis. A novel image enhancement approach and a conclusive image registration approach were proposed in this paper.The method based on wavelet transform of image enhancement, may adopt different forms of treatment to the different scale and different level of frequency .it takes more flexibly ,effectively . it can also enhance the details of image , weak the noise of image. Consequently, this method was an effective medical image enhancement. The registration based on wavelet transform, mainly make fully use of the different feature space, similarity metric and search strategy of different scales, find the best result of registration. Because the number of pixels of different scale and resolution, we begin with the lowest resolution to find the firstling result. along with the amount increased , the firstling will impact on the accuracy of the follow.The image enhancement based on wavelet transform, Firstly, the inverse distance weighted average method was used to reduce a medical image's noise. Secondly, the medical image was decomposed with wavelet transform, and at the same time, all high-frequency sub-images were decomposed with Haar transform. Thirdly, high-frequency coefficients were enhanced by different weight values in different sub-images, and the low-frequency sub-image is multiplied by a factor which value is less than 1. It can help to highlight high frequency information effectively. Then, we get the initial enhanced image through the inverse wavelet transform and inverse Haar transform. Finally, the initial enhanced image's histogram was stretched by using nonlinear histogram equalization, and then the last enhanced image is obtained. Experiments showed that this method not only can enhance an image's details but also can preserve its edge features effectively.The medical image registration based on wavelet transform which used the enhanced medical image. Firstly, reference image and floating image were decomposed by wavelet transform. Secondly,?to extract the feature of the reference image, we chose some feature points from the lowest-frequency and proximate sub-image as the initial data. Thirdly, required feature point mapping from the tow images which are the lowest-frequency reference image and the lowest-frequency floating image, using the algorithm of Mutual information and improved genetic algorithm. Then got the angle, horizontal translational variable, vertical translational variable. Fourthly, took the three variable quantities to the next higher lever of frequency until the highest resolution, received the best feature point. Finally, solved the parameters of rigid transform model through the last three quantities.