X-ray medical imaging. 2D image processing

The image delivered by an X-ray detector is unusable as it stands. First of all, it extends over a dynamic range of gray levels far greater than the eye can detect, and far greater than the display screen can display: this dynamic range must therefore be compressed, while retaining perceptible contrasts of interest. What's more, every detector must be calibrated to render the effects of sensitivity (or gain) fluctuation invisible, and also to obtain an image with impeccable black (offset suppression). Any defects in the detector (dead pixels, etc.) must also be made invisible. All these operations result in a "clean" image, which is still not what the radiologist expects: it has to be filtered to highlight one tissue rather than another, one detail rather than another, depending on the nature of the examination and the patient's body characteristics. This type of processing has developed considerably since 1995, following the advent of digital detectors (CR cassettes, CCD cameras and, more recently, flat panel detectors), and also thanks to the commercial availability of increasingly powerful processors.

This article looks at the mathematical processing of X-ray detector images. Its aim is not to give a theoretical description, which would be too rigorous. Neither will the reader find precise algorithms, but he or she will find a physical description and an explanation of the objectives sought.