Super-resolution Analysis On Millimeter-wave Radiometric Imaging

Millimeter-wave (MMW) is the electromagnetic wave between microwave and infrared ray. MMW detective technology has well directivity, strong anti-interference ability, good detection performance, and could distinguish metal targets and environment commendably. MMW radiometric imaging is a MMW detective technology that could obtain abundant information about intuitionistic configuration of objects. Consequently, millimeter-wave imaging has become a research focus in recent years.Millimeter-wave imaging systems have problems in low spatial resolution and poor sensitivity. As millimeter-wave imaging technology being used more and more extensively, the requirement with quality of millimeter-wave images has enhanced. Super-resolution algorithms could resume the information out of cut-off frequency to get higher spatial resolution and plain image under the condition of present hardware. The research of super resolution for millimeter-wave image is valuable.This paper introduces the development and application prospect of millimeter-wave imaging technology, and summarizes the research status of super-resolution algorithms firstly. Then analyzes the basic principle of millimeter-wave imaging and discusses the reasons for image degeneration in detail. Some super-resolution algorithms based on spatial domain are applied in more extensive area because of including more prior constraints of spatial domain. Discusses certain popular algorithms based on spatial domain; analyzes different effects, superiorities and restrictions of different methods.Based on the model of millimeter-wave imaging and the analysis for different super-resolution algorithms, an improved method is presented based on projection onto convex sets (POCS) algorithm. The new algorithm filters low resolution image to de-noise additive noises mentioned in the imaging model and get a meliorative image. In order to get an initial image for POCS amending process, interpolates the meliorative image with improved wavelet bi-linear interpolation. POCS amending process is based on point spread function (PSF) and statistical characteristic of noises. For every high resolution image pixel influenced by point spread functions of several low resolution image pixels, uses the mean of all amended results as the final amended result in every circulation, instead of amending immediately, so that reduce the effect on anterior amendments.Experiment results show that this method is better than conventional interpolations, wavelet bi-linear interpolation and improved wavelet bi-linear interpolation. The new algorithm could enhance peak signal to noise ratio (PSNR) of super resolution images effectively, and these images have well visual quality. For MMW images, the proposed method could obtain good results.