Research On Bayer Image Restoration And MTF Test Technology

With the development and application of imaging technology,color image has already became an important carrier of information and played an important role in many scientific research and detection activities.Color image processing research has highlighted its important application value.Color camera covered with color filter array is one of the most common and practical imaging methods.Its data amount is as much as grayscale image,and color information can be reproduced after interpolation.It is a widely used method of color camera at present.In this paper,the related technology of color filter array color camera is studied.From the perspective of application scene analysis,due to the difference of optical lens,filter parameters,sensor parameters and restoration algorithm of color imaging camera,the color of images collected and restored by the camera will deviate.In order to obtain high-quality images,color correction is often needed.The image quality of color camera also needs to be evaluated.In engineering,the image quality of space camera often needs to be monitored in orbit.In the research of color correction using standard color card,it is necessary to establish the mapping between the standard color and the measured camera color.At present,most of the research focuses on improving the mapping accuracy.However,in the undetermined problem of using 24-color correction for the whole color domain,how to improve the smoothing term of the mapping is an issue to be discussed.In the evaluation of imaging quality by using Modulation Transfer Function,the most commonly used method is to use the edge method to measure the Modulation Transfer Function(MTF).However,due to the unique sampling characteristics of color cameras,the edge method is not accurate enough to accurately measure.In this paper,the color correction of Bayer mode color camera and MTF test technology are studied.For the part of color correction,on the one hand,the improvement of correction accuracy is discussed.On the other hand,the generalization performance of color mapping is discussed.This article chose two kinds of algorithms,respectively is automatically piecewise polynomial algorithm and neural network algorithm with constraints,create a mapping for calculation and color card,then the calibration accuracy and reasonable assessment,piecewise polynomial algorithm to calculate color difference can reach 4.99,the color difference value is less than 6 in the CIELab uniform color space,color correction precision in the acceptable range.The color difference of BP neural network can reach 3.06,the color difference value is close to 3,and the visual perception is almost imperceptible.Finally,the calculated mapping was used to calibrate the real image,and the correction performance was evaluated by subjective evaluation and color parameters.The mean and standard deviation of the images before and after the correction were calculated to evaluate the color variation characteristics.The simulation results show that the neural network algorithm is superior to the automatic piecewise polynomial algorithm,both of which are superior to the traditional polynomial algorithm in terms of calculation accuracy and generalization performance.For the testing part of color camera's modulation transfer function,the sampling characteristics of camera's detector in Bayer format and the cascade of camera's modulation transfer function are analyzed.The error caused by direct calculation without channel differentiation is discussed.The modulation transfer function of RGB channel is calculated by channel weighting.The Edge Spread Function(ESF)and Line Spread Function(LSF)were further solved by using a 3D surface fitting method to fit the sparse sampling of 3D surface.Finally,the weighting factor of three channels is calculated and the overall MTF is weighted.The calculated results are close to the static MTF design values of the imaging camera.