Research And Application Of Sub Pixel Registration Algorithm For The Sun Image

In the observation of solar, usually need to accumulate multiple frames short exposure images to obtain high SNR images. Linear displacement is exist due to influence of atmospheric turbulence and wind; In that case, image registration techniques should be applied. With the improvement of the accuracy of image registration, the accuracy of image registration requirements sub-pixel level, therefore, integer pixel image registration algorithm can not meet the needs of solar observation, while the traditional sub-pixel image registration algorithm exists interpolation error, large amount of computation and so on, with the high-performance CCD widely used in astronomical telescopes. The image acquisition speed is becoming faster and faster. So the image registration needs to be done in a short period of time, proposed a huge challenge to the image registration algorithm, and the astronomers.The main work and results are as follows:First, analyze the existing problems of the traditional sub-pixel image registration, base on the cross-correlation, proposed and implemented sub-pixel image registration algorithm based on the modified moments.Second, by comparing measurement accuracy and anti-noise performance of the subpixel image registration algorithm based on the modified moments and the classical algorithm on. Processing both the simulated and measured images, under different noise level, in the case of low signal-to-noise ratio, the situation of subpixel image registration algorithm based on the modified moments is obviously superior to the Guizar-Sicairos algorithm, next to Keren algorithm. And the speed is orders of magnitude improvement, can reach high accuracy in the high SNR, although under the high SNR condition the Guizar-Sicairos algorithm can achieve higher accuracy, but the registration speed slower.Third, the subpixel image registration algorithm based on the modified moments applied to local correlation tracking. By comparing and analyzing the correlation tracking image of phase correlation and cross-correlation, the correlation tracking image of the subpixel image registration algorithm based on the modified moments display jitter of the local area. Provide the local image jitter, atmospheric jitter information for astronomers.Fourth, we use different registration algorithms to evaluate the accuracy of the image of solar magnetic field. Processing the measured images, we choose five statistics to evaluate the results of the processing, which are the mean and variance of the difference images, the maximum correlation value of registration images, entropy of accumulation images, and magnetic energy of the images. We conclude that the phase correlation algorithm is more sensitive to noise than cross correlation, and cross correlation algorithm is faster than the phase correlation according to the algorithm principle, so the cross correlation algorithm is more suitable for the real time processing of solar magnetic field images, image of the solar magnetic field obtain by subpixel image registration algorithm based on the modified moments with high resolution.For the sun image, it does not have a clear target contour and the SNR is lower than ordinary image, the subpixel image registration algorithm based on the modified moment performance better on anti-noise, measurement precision and speed.