A linear interpolation method in a CT image reconstruction back projection process

Abstract

The invention relates to a linear interpolation method in a CT image reconstruction back projection process. As for pixels, defining a radius-variable disc taking the pixel point P as the center of acircle; selecting a radius coefficient according to the image view field and the pixel point number; enabling the circular disc to meet a resolution requirement and a data processing capability, calculating and obtaining a channel boundary covered after the circular disc is projected to a detector, performing weighted average on image values of all channels covered after the circular disc is projected to the detector to serve as values of pixel points, and enabling a weighting coefficient to be according to a coverage ratio of a corresponding channel. The method has good signal-to-noise ratioand resolution, has good elasticity and adaptability for different situations, when the pixel size becomes small, the interpolation method is degraded to be approximate to a linear interpolation method, it is guaranteed that the better signal-to-noise ratio is achieved when a large matrix is used for reconstructing a CT image, and the method can be better suitable for development of the CT image reconstruction technology.

Description

Technical field [0001] The invention relates to a linear interpolation method in the process of CT image reconstruction and back projection. Background technique [0002] Back projection is one of the main processes in the CT image reconstruction algorithm. The existing back projection methods are roughly divided into two categories, one is the ray-drivenback projection method, and the other is based on The pixel-driven back projection method (pixel-driven back projection), but in practice, considering the simplicity and ease of implementation, the pixel-based back projection method is mostly used. [0003] The traditional pixel-based backprojection method assumes that a given pixel has no size, but a pure point. Considering that the intersection of the x-light source point and the given pixel point P to the detector is usually between two Between two detector units, in order to obtain the pixel corresponding to the data on the detector, it is usually necessary to use an interpola...

AI Technical Summary

Problems solved by technology

[0033] Compared with other existing technologies, the image reconstruction method of US Patent 8,116,426 regards pixels as discs whose radius is half the size of the grid. The implementation process is relatively simple and easy to use, but the pixel model is not flexible enough

First of all, this method regards the pixel as a disc whose radius is half the size of the grid. Once the grid size is fixed, the radius of the disc is also fixed (see formula (2)), which lacks elasticity, so it has a larger field of view. When the number of pixels is relatively constant, the pixel size is too large. Because the pixel size is too large, the number of channels covered by the projection onto the detector is too large, which will cause the reconstructed image to have low noise but high resolution. If the ratio is too low, the details will be lost; secondly, the method will degenerate into the nearest neighbor interpolation method when the grid size becomes smaller, which can be seen from the limit case ΔS=0. When ΔS=0, formula (5) and ( 6) Given gamma i = γ j = γ, so that c i =c=c j , and from formulas (4) and (15), it can be seen that this is actually the nearest neighbor interpolation method, so when the pixel size becomes smaller, the image obtained by this method will have larger noise

There are at least two situations where the pixel size becomes smaller. From the formula (2), we can see that the first case is that if the number of pixels N of the image remains unchanged, the smaller the field of view FOV will cause the pixel size to become smaller; the second case The second situation is that if the field of view size FOV remains unchanged, and the number of pixels N becomes larger, the pixel size will also become smaller, and the second situation is currently becoming a trend in CT image reconstruction, that is, using a large matrix (such as 1K or 2K matrix) to reconstruct CT images, therefore, US Patent 8,116,426 is difficult to adapt to the development requirements of CT image reconstruction technology

Images