Image Processing Research For 2×2 Optic Tapers Array X-ray Detector

Compared with traditional absorption imaging,X-ray phase contrast imaging is a new imaging method since it obtains contrast of images by recording the phase distribution of different objects or different parts of one object which X-ray pass through. It is a revolutionary breakthrough in the field of X-ray imaging because it resolves the problem fundamentally that traditional absorption imaging has a poor contrast when applied on objects of light-element.One of the problems in studying X-ray phase contrast imaging system is the technology of X-ray detector with big area. In order to using those existing X-ray detector technology, we match those smaller ones coupled with CMOS Chips to get detectors with appropriate area for examination of human organs. Affected by the imaging characteristics of fiber optic taper array, the final image has kinds of problems such as geometric distortion, absonant brightness and so on. In fact, the mechanical coupling of fiber optic taper array in the current version is the primary cause leads to image geometric distortion. To solve these problems, this article designs corresponding solution,and the primary work is as follows:(1) Geometric correction of single fiber optic taper(FOT): The use of FOTs results in geometric distortion in the final image. According to experiments, the distortion in our study is pincushion distortion. This article uses a distortion correction algorithm based on the polynomial fitting. In this method, we build the corresponding relation between desired image and distorted image, and get the correct gray value of a point in desired image according to its corresponding point in distorted image, and the degree of geometric distortion is greatly improved.(2) Image stitch of four FOTs: Only the middle part of original image from the system is effective due to the fact that the area of the small end of the optic taper is smaller than that of the CMOS chip coupled with the taper. To deal with images in different places, we turn over these images in different ways, and find its boundary to get the effective part;(3) Noise processing: the image noise consists of two parts: hardware system noise and background noise. The former could be smoothed according to its own characteristics, but given that direct processing for background noise may fuzz the details up, and reduce the resolution, this article come up with a modified median filter algorithm to protect edge information effectively.