The computational method for brightness alignment of digitalimages

Abstract

The purpose of this work is the alignment of the interference substrate on the large CCD frames without using of service (Master) frames (dark current frames – dark, noise reading frames – bias, «flat field» frames – flat). Some cases when it is impossible to get Master-frames or calibration of received image don't lead to the desired result, for example, in the presence of ambient light. The images of celestial objects (stars and asteroids) on a separate frame are point, and with the atmosphere turbulence, can be presented as «blurred points». This background unevenness is presented as slowly varying brightness changes. Therefore, it seems appropriate to consider the frames processing in the frequency range, where the stars and asteroids are formed by high-frequency spectral components of the image and the background is the low-frequency components. In this case, in order to remove lowfrequency background variations and to leave the unchanged high-frequency components of the image, it is necessary to perform high-pass filtering for the image. Median filtering is a method for nonlinear image processing. It is effective if the pulsed noise is the limited set of peak values against of zeros. The median filter is implemented as a local processing procedure by sliding window with specified sizes that includes an odd number of image counts. The processing procedure consists of the readings which trapped in each position of the window, ordered by ascending values. The average count in the ordered list is called the median of this group. This median replaces the central count in window for the processed image. Thus a median filter suppresses the pulsed emission of the original image, if the area of the pulse signal is less than half of the filter’s aperture area. Therefore, the frame can be obtained with only large structural components of the background after the frame filtering with median filter of appropriate size. Background unevenness will be compensated during subtracting it from the original frame. Results of median and frequency filtering are comparable. They change the medium and standard deviation of the histogram of the frame background. Also these filters increase twice signal and noise ratio (SNR) of objects images, including dim. The examples of aligned images are provided.