Relative Feature Distance Based Region Matching Interpolation Of Sectional Image

In the medical figure processing, people often hope to reconstruct the three-dimensional objects, set up virtual organs and tissues for medical diagnostic and calculate the radiation treatment planning in three-dimensional dose field under three-dimensional tomography images. However, the resolution required of the image is unknown when image is gained, the limitations of image acquisition devices and losing and so on, the resolution obtained during image acquisition and the resolution of image processing are often repugnant, the distance between the images is greater than the cross-sectional image distance between pixels on the image. In order to carry out three-dimensional reconstruction, image interpolation method is proposed to improve the resolution between layers, the image interpolation method is used in the original cross-sectional images and then generates a number of intermediate inter-sectional images so that the image resolution between layers and cross-sectional image distance between the pixels are unanimous.Scene-based interpolation method is a image interpolation algorithms using the image gray level information in the form of constructing the interpolation curve, it is simple and easy to achieve, but whether linear interpolation or higher-order non-linear interpolation, they are a smooth processing of the known data, just a low-pass processing essentially, the boundaries of the object mainly decided by the high-frequency components,so it will lead to blurred contours and boundaries ghosting of interpolated images definitely and have a very large difference with the real images. Object-based interpolation method is also known as shape-based interpolation algorithm, for the traditional gray-level interpolation method will lead to blurred contours and boundaries ghosting of interpolated images and it is not conducive to the extraction of the image border and further image processing, shape-based interpolation method using the shape information of the CT images to produce the middle contour of the interpolated images and using the middle contour to control the image interpolation, blurred contours problem can be better solved and it is facilitate for the image display and further image processing, but the shape-based interpolation method depends on the image shape information extraction heavily and each interpolation can only get a shape one time, for complex images it is difficult to extract the shape information, and it is powerless for the border of the internal organs, the method has a bad result for the iamges with relatively large difference. Wavelet-based interpolation method can get a series of wavelet sub-images and low-frequency subgraph through wavelet transform, and then interpolate between the wavelet coefficients of the same scale and orientation of the wavelet sub-images, Finally the new wavelet sub-images and low-frequency sub-graph can be got by inverse wavelet transform, then obtain new interpolated images, this method interpolate with the gray and shape information of the original images at the same time and fully simulate the middle physical changes in the morphology from one image to another image and produce high-quality images, but the large amount of calculation is not conducive to practical application.To address the above problems, in this paper, a new method is presented:First, we build up the corresponding relation between the interpolated images by an optimization process which was based on the gray value, the gradient of the gray value, the gradient direction in neighboring slices, the distance of the pair point and the gray-scale continuous threshold we have set to get the best paired points; second, for the best-matching point pair, linear interpolation, and for the worst-matching point pair, we calculate the shortest feature distance to the relative feature points and use the shortest feature distance to control the gray-scale interpolation of the corresponding.point.Finally, the new method is used to interpolate images. Compared with existing approximate method, our method can produce better result.