Research And Implementation Of Medical Three Dimensional Reconstruction System Based On GPU Parallel Computing

In the past, medical diagnostic imaging simply picks up information from the collection to the site of lesions observed in patients. Because it cannot build stereoscopic vision of observation, its diagnostic results often bring with doctors’ subjective judgments. The patient of health and even life which lies in the doctor’s clinical experience, is not scientific and accurate. However, the three-dimensional reconstruction of medical images can be extracted from the two-dimensional information to model three-dimensional images and make medical imaging diagnosis be observed from stereo direction, reducing the subjective error of doctors. It has greatly enhanced the accuracy of diagnosis. While traditional three-dimensional reconstruction requires enormous amount of data and computation, resulting in running slowly and not easy to transform, so it is not widly used. This article in this context puts forward improvement of three-dimensional reconstruction way, first to two-dimensional image preprocessing,the GPU parallel computing way of3D reconstruction.Because the medical tomographic CT image’s clarity is not enough and vulnerable to noise influenced trait, if it uses original image3D reconstruction, the effect is always not very good. So the three-dimensional image preprocessing is necessary. However, the traditional denoising models in removing noise often destroys image boundary, resulting in image boundary not clear. This paper is proposed based on anisotropic partial differential denoising models, can save image boundary as well while denoising, and receive good treatment effect.Improving the efficiency of3D reconstruction is an important research direction in the medical visual field. Traditional methods to improve the image quality is often under the premise of the expense of reducing the number of generated patches,or is generated by running special image accelerate card. Due to high cost or low accuracy, it also be used widely. This paper introduced a way of using GPU powerful parallel computation ability to transform MC algorithm’s the original serial operation method to parallel processing, which greatly improves the efficiency of the algorithm operations.