The current infrared detection systems commonly memory and transfer image signals in the form of video. However, when the videos are in the process of formation, transmission and storage, they are easily polluted by motion blur and noise. Accordingly, the video motion blur recovery algorithm was proposed based on this system. Firstly, the video motion blur restoration module was built based on video streaming by integrating mutual information of every frame of sequence images. Secondly, the corresponding algorithm was put forward and the point spread function (PSF) was estimated effectively. Thirdly, the motion blur recovery process was described and all the function module were created. And then, in order to reduce the calculation burden, the image sequence was equal interval sampled from the original video, which enhancing the image quality and achieving better restoration effect. Finally, a subjective and an objective evaluation system were introduced to compare our algorithm with two other classical algorithms and evaluate results. The experimental results show that the peak signal-to-noise ratio of each frame of restored video reached 37, mean square error was below 9, which was superior to the control algorithm. The results basically meet the requirements of detection system, which discovering targets and monitoring the airspace.
As a new imaging mechanism, ghost imaging has become a hot area of research in optical imaging field. In this paper, the effect of intensity correlation order on lensless ghost imaging system is investigated. We demonstrate that the image quality of Nth-order ghost imaging and Nth-order ghost imaging with background subtraction can be affected by the different intensity correlation order of test light and reference both theoretically and experimentally. The result indicates that the image quality will not be certainly increased with the increasing of the intensity correlation order, and here will be very useful for choosing an appropriate intensity correlation order in practice.
The goal of infrared image restoration is to reconstruct an original scene from a degraded observation. The restoration
process in the application of infrared wavelengths, however, still has numerous research possibilities. In order to give
people a comprehensive knowledge of infrared image restoration, the degradation factors divided into two major
categories of noise and blur. Many kinds of infrared image restoration method were overviewed. Mathematical
background and theoretical basis of infrared image restoration technology, and the limitations or insufficiency of existing
methods were discussed. After the survey, the direction and prospects of infrared image restoration technology for the
future development were forecast and put forward.
Proc. SPIE. 7872, Parallel Processing for Imaging Applications
KEYWORDS: Infrared search and track, Infrared imaging, Digital signal processing, Logic, Detection and tracking algorithms, Data modeling, Data storage, Field programmable gate arrays, Infrared radiation, Electronics engineering
The paper presents a low cost FPGA based solution for a real-time infrared small target tracking system. A specialized
architecture is presented based on a soft RISC processor capable of running kernel based mean shift tracking algorithm.
Mean shift tracking algorithm is realized in NIOS II soft-core with SOPC (System on a Programmable Chip) technology.
Though mean shift algorithm is widely used for target tracking, the original mean shift algorithm can not be directly used
for infrared small target tracking. As infrared small target only has intensity information, so an improved mean shift
algorithm is presented in this paper. How to describe target will determine whether target can be tracked by mean shift
algorithm. Because color target can be tracked well by mean shift algorithm, imitating color image expression, spatial
component and temporal component are advanced to describe target, which forms pseudo-color image. In order to
improve the processing speed parallel technology and pipeline technology are taken. Two RAM are taken to stored
images separately by ping-pong technology. A FLASH is used to store mass temp data. The experimental results show
that infrared small target is tracked stably in complicated background.
Background estimation plays an essential role in many infrared (IR) target detection algorithms. A kernel-based
background estimation algorithm for stationary camera is proposed in this paper. The nonlinear version of least squares
(LS) algorithm: kernel least squares (KLS) and its exponential weighted form (KEWLS) are deduced use kernel methods
(KMs). The background of IR image is estimated by KLS or KEWLS nonlinear regression utilize sequence images as
training set; then targets are segmented by threshold dependent techniques in the difference image. Experiments of
nonlinear function regression and IR image background estimation are performed. The results of these experiments are
compared to that of LS algorithm, a single-frame and a multi-frame background estimation algorithm. The feasibility of
nonlinear function regression and background estimation via kernelized LS is thus demonstrated.
The desire to maximize target detection range focuses attention on algorithms for detecting and tracking point targets.
However, point target detection and tracking is a challenging task for two difficulties: the one is targets occupying only a
few pixels or less in the complex noise and background clutter; the other is the requirement of computational load for
real-time applications. Temporal signal processing algorithms offer superior clutter rejection to that of the standard
spatial processing approaches. In this paper, the traditional single frame algorithm based on the background prediction is
improved to consecutive multi-frames exponentially weighted recursive least squared (EWRLS) algorithm. Farther, the
dual solution of EWRLS (DEWLS) is deduced to reduce the computational burden. DEWLS algorithm only uses the
inner product of the points pair in training set. The predict result is given directly without compute any middle variable.
Experimental results show that the RLS filter can largely increase the signal to noise ratio (SNR) of images; it has the
best detection performance than other mentioned algorithms; moving targets can be detected within 2 or 3 frames with
lower false alarm. Moreover, whit the dual solution improvement, the computational efficiency is enhanced over 41% to
the EWRLS algorithm.