Effective Photon Count Estimation from a Single Image Acquisition

Most image sensors do not directly report the actual number of detected photons per pixel, but instead a value generated after the photo-electrons registered on a physical sensor have passed through an electronic pipeline of amplifiers and charge offsets. Knowledge of the actual photon count is necessary to implement maximum likelihood estimates to extract model parameters from biological fluorescence images. In order to estimate the photon count of an image, the effective gain and offset of the image acquisition device must be measured. The standard gain calibration process for cameras is performed by recording tens of calibration and dark images and then performing a linear regression of the gain by plotting the variance over the acquisitions as a function of intensity. We propose a method to compute the gain from a single image acquisition without pre-calibration by utilizing information in the frequencies outside of the optical transfer function (OTF). Since noise is not limited by the cut off frequency of the OTF, estimation of the out of band energy relative to the total energy allows for the calculation of the gain. We show that this procedure is as precise as traditional gain estimation and is an essential step in automating super resolution software.