Research On Normalization And Registration For Biomedical Image

The microscopic repair of peripheral nerve injury has been a hot and difficult part in the area of surgical clinical research. The development of computer technology and pattern recognition technology, made it possible to achieve3d reconstruction of peripheral nerve function bundle.3d reconstruction of peripheral nerves mainly consist of the follow steps:Two-dimensional image acquisition, image preprocessing, normalized, segmentation, registration, recognition and3d reconstruction, etc. The purpose of this paper is to explore a suitable normalization and registration method for Peripheral nerve slice images. Based on image normalization processing, all forms of slice images has been put into its normal form and the generality of the follow-up image processing algorithm has been improved; the process of image registration Lay a foundation for3d reconstruction of peripheral nerve slice image.the main work in this paper is as follows:1. First of all, this article simply introduces the2d image acquisition process, and on the basis of that, analysis noise that may attached. Pointed out that since sliced need artificial dyeing\micrography, etc, so even for the same slice, the resolution size, rotation Angle and gray may be different, In order to eliminate the differences, Image normalized and registration is necessary. In addition, The vector median filter used in image filtering, it improved image quality.2. Secondly, based on the analysis of the advantages and disadvantages of the invariant moment, proposed a spatial normalization method that using Pseudo-Jacobi-Fourier Moments. This method can make the image as the only standard and the standard image has invariance for affine transformation, then, proposed gray scale normalization method that using histogram to improve image gray feature.3. The third, Method of image registration based on mutual information is proposed, This method takes all the pixel gray as feature space, mutual information as similarity measure, Using Powell optimization algorithm for parameter optimization and get the optimal parameters for space transformation. In the experiments, Space normalized transformation parameters can be the initial transform parameters.