3D Noise Computation Of Infrared Imaging System Based On Finite Sampling

In recent years,infrared thermal imaging systems have been widely used in various industries.Infrared thermal imaging systems shows its advantages from the military field to other industries.People demand more about requirements for the performance and analysis methods of infrared imaging systems.Image noise is one of the important factors that affect the infrared imaging system.Modeling and evaluating the infrared imaging system is very important for it's optimization and development.This paper firstly summarizes the main noise and the noise model,briefly describes the calculation method of the three-dimensional noise model,and discusses the statistical bias in the classical method.This paper mainly optimizes the three-dimensional noise calculation method for finite sampling.By obtaining the unbiased sample variance in each direction and finding the coupling relationship with the true variance,seven unbiased three-dimensional noise items are obtained.Since each calculated three-dimensional noise term represents an estimated value of the noise variance calculated from a total random sampling,this paper derives the confidence interval for each three-dimensional noise variance by calculating the variance and mean of the three-dimensional noise variance estimation.The confidence interval can give the possibility that the true value is included within this range within a certain probability,and quantitatively give the credibility of the calculated three-dimensional noise variance.Secondly,the paper also discusses the three-dimensional noise calculation method that minimizes the required samples within the accuracy range of the confidence interval.This method constructs a sub cube to perform variance calculation by spatial sub sampling.It reduces the need for the maximum number of samples within the predicted range of the derived confidence interval,and can evaluate the range of variation of imaging system possibly observed in the imaging space from two-dimensional images.The method provides a better data observation model for the analysis of three-dimensional noise.At the same time,Monte Carlo simulation is used to verify the feasibility of improving the accuracy of variance calculation under finite samples.A simulation scheme is designed which generates a three-dimensional noise cube by setting the initial value of the three-dimensional noise term to calculate the three-dimensional noise of the noise cube.The calculation result is compared with the initial value to obtain the calculation deviation.By comparing the deviation percentages of mean deviations from the initial values of Monte Carlo simulations after different Monte Carlo simulations for different methods,it is verified that our method does significantly improve the results of the three-dimensional noise calculation under finite samples.Finally,we measured the three-dimensional noise of the 35°C and 50°C blackbody images taken with a 240×320 long-wavelength infrared camera.The results are basically consistent with the predictions.The research in this paper provides a more ideal data observation model for the analysis of three-dimensional noise model and quantifies the source of noise,witch provides a theoretical basis for the improvement of detector structure and performance improvement and new ideas for the performance evaluation of infrared thermal imaging systems.