Study On Application Of Wavelet In3-D Reconstruction Of CT Images

With the rapid development of digital imaging technology for medical, advanced medical imaging equipments for example CT, ultrasound, NMRI, SPECT provide high-quality2D images of organs, bones, joints and so on. In the diagnosis, combining acquired2D images to form3-D model will provide high reliability, more fully zone information. It will help doctors deeply understand illness and raise the treatment rates.This paper took the wavelet analysis as research background and did some research on interpolation algorithm between two CT images.3D brain models have been done through combing interpolation images and original CT images. The main contents are as following:Firstly, this paper researched wavelet analysis especially its multi-resolution analysis theory. A image was been processed by using second wavelet transform. We knew that low frequency subgraphs included most of all energy but little detail and high frequency subgraphs included most of all details by analyzing the subgraphs after wavelet transforming the image and energy diagram. According this property, we can manipulate different frequency subgraphs images using different methods.Secondly, the method of interpolation based on SIFT matching was been put. Experimental results shew that the method overcome the overlapping problem and edges and details were not clear on the transition zone of different gray value. The effect was superior to scene-based and object-based. So this method is feasible.Thirdly, this paper present a new method of combining the wavelet analysis and interpolation based on SIFT matching based on the requirement of clear edge.The process and experimental results of this method were been showed. CT images were processed by wavelet transforming, then interpolation was done between high frequency subgraphs by this new method and interpolation was done between high frequency subgraphs by scene-based. This new method had higher PSNR and less unequal gray value points. The interpolation image had a natural transition zone, clear edge and rich details.Finally,3-D reconstructions based on treated2D serial sections were performed. And surface rendering was compared with volume rendering combining our practical study. The result showed that the volume model concluded rich organs information, rendering visual effects and guided to diagnose the illnesses better. It is significant to diagnosing the disease.