Research of liquid CT image de-noising based on improved NL-means algorithm

The terrorist incidents in recent years have promoted the computed tomography(CT) technologies for liquid explosives detection in civil aviation. The motivation of X-ray CT application for liquid explosive detection is to identify and classify liquids filled in bottles without opening. In the liquid computed tomography system for security inspection, because of the presence of all kinds of noises such as quantum noise, electron noise and so on, the system performance and liquid CT image quality are degraded so as to influence the statistical calculation of CT number. Compared with the current main de-noising methods, non-local means(NL-Means) de-noising algorithm proposed by Buades provides the advantage of convenience in design for implementation, which estimates noise-free pixel intensity as a weighted average of all pixel in the image and weights proportionally to the similarity between the pixel being proposed and its local neighborhood items. Unfortunately, it is prone to produce "staircasing effect" on image having nonzero tone gradients. In this paper an improved NL-Means algorithm was proposed in detail, which mainly changed de-noising key role of the weighted kernel function. The experiment result indicated that the improved NL-Means de-noising algorithm could suppress CT image noise effectively and simultaneously preserved the spatial resolution.