Fast measurement of temporal noise of digital camera's photosensors

Abstract

Currently photo- and videocameras are widespread parts of both scientific experimental setups and consumer applications. They are used in optics, radiophysics, astrophotography, chemistry, and other various fields of science and technology such as control systems and video-surveillance monitoring. One of the main information limitations of photoand videocameras are noises of photosensor pixels. Camera’s photosensor noise can be divided into random and pattern components. Temporal noise includes random noise component while spatial noise includes pattern noise component. Spatial part usually several times lower in magnitude than temporal. At first approximation spatial noises might be neglected. Earlier we proposed modification of the automatic segmentation of non-uniform targets (ASNT) method for measurement of temporal noise of photo- and videocameras. Only two frames are sufficient for noise measurement with the modified method. In result, proposed ASNT modification should allow fast and accurate measurement of temporal noise. In this paper, we estimated light and dark temporal noises of four cameras of different types using the modified ASNT method with only several frames. These cameras are: consumer photocamera Canon EOS 400D (CMOS, 10.1 MP, 12 bit ADC), scientific camera MegaPlus II ES11000 (CCD, 10.7 MP, 12 bit ADC), industrial camera PixeLink PLB781F (CMOS, 6.6 MP, 10 bit ADC) and video-surveillance camera Watec LCL-902C (CCD, 0.47 MP, external 8 bit ADC). Experimental dependencies of temporal noise on signal value are in good agreement with fitted curves based on a Poisson distribution excluding areas near saturation. We measured elapsed time for processing of shots used for temporal noise estimation. The results demonstrate the possibility of fast obtaining of dependency of camera full temporal noise on signal value with the proposed ASNT modification.