The Study On Fast Texture Improving Algorithm For Ray Imaging

According to the form of radiation, ray can be divided into two types, one iselectromagnetic radiation, its quantum is photons, such as gamma rays and x-rays.One is particle radiation, such as all kinds of particle radiation, alpha particles, betaparticles, protons, electrons, neutrons and so on. X-ray imaging is widely used inmedical tests and industrial detection，due to its ability that can show the subtleinternal structure. There is time and ray intensity limit about the x-ray nondestructivetesting in practice, which leads to obtained image resolution not enough. Therefore,the image processing method must assist detection task rapidly; Local characteristicsof image are embodied in gray contrast and disease, that requires the correspondingimage processing method has remarkable effect on image texture feature extracting.As known，there always exist severely degradation problems in digital radiography.How we can extract necessary textures from degraded radiographic images is thepost-processing key.In order to make detection more quickly, LBP(local binary pattern) is used fordegraded ray image texture extraction. It can describe the image texture informationwith the method of weighted and the binary number, that make the circle domainpixel gray level difference reflected very well. LBP has three advantages: thecalculation speed, drab gray scale invariance and information extraction of locality.Local binary pattern is a well-known method, which is widely used in fast imagetexture extraction. However for noisy images, LBP can’t work well due to itssensitivity to details. On the other hand, as one of the important shock filtersdeveloped in recent years, complex shock filter possesses excellent capabilities intextural image processing. Here, by combining complex shock filter with LBP, anovel fast and efficient method, C-LBP is presented for texture extraction ofdegraded radiographic images. Firstly, we introduce complex shock filter for imagepreprocessing, then take the imaginary part image data as the input for next textureextraction as the higher stability of imaginary part data and the stronger antinoiseability. Secondly, embed a comparative smooth comparison function to improve theLBP operator from aspects of the weighted method and the comparison functionconsidering round similar distance, neighboring points and the center of intensityand gray information to differentiate between circle domain. Finally, a counting scheme is utilized in order to eliminate redundant tiny gray level change.Experimental results show that comparing with traditional LBP, C-LBP not onlydistinguishes between noise and details in radiographic images, but also extractsimage textures efficiently and rapidly, which plays an important role in developingnondestructive detection technique by low-dose ray radiography.