Comparison of EMCCD post-processing methods for photon counting flux ranges

ABSTRACT

Several post-processing methods were proposed to overcome the excess noise factor induced by EMCCD multiplication register. Each method has a unique effect on SNR. However, since SNR does not account for photometric accuracy, it cannot be reliably used to directly compare the performance of these algorithms. A normalized quadratic error that accounts for both SNR and accuracy is proposed as an alternative figure of merit. This approach provides a quantitative and rigorous comparison. Using both experimental and simulated frames in the faint-flux range, it is used to compare the existing EMCCD post-processing methods

1. INTRODUCTION

Charge-Coupled Devices (CCDs) are widely used for imaging and their sensitivity is acceptable for most applications. However, when light become sparse, which is the case for high frame rate or low lighting applications, the signal-to-noise ratio (SNR) of these devices drops rapidly due to their readout noise. For these faint-flux imaging applications, overcoming the readout noise is essential. The advent of the EMCCD made it possible to break readout noise barrier, rendering single photon detection possible on a CCD focal plane.1

This low readout noise comes at the price of an Excess Noise Factor (ENF) that, from a SNR point of view, has the same effect as if the quantum efficiency of the device were halved.2 Analog Readout (AR), where the signal of the pixel is calculated by simply dividing the output signal by the mean gain of the EM register, is plagued by this reduction in SNR. Authors have proposed methods to circumvent the ENF (Photon Counting,1, 3–6 Thresholding,7 Multiple Thresholding,4 Bayesian Inference7 and Thresholding with Bayesian Inference8 ). By taking into account the main sources of noise affecting the EMCCD (dark current, clock-induced charged, and readout noise), they aimed at best estimating the true input flux on the CCD from the output signal. This paper reviews and compares these post-processing methods.

Since the purpose of these post-processing methods is to most accurately deduce the real input photon flux on a pixel from the digital readout, a proper figure of merit must be chosen to yield a coherent comparison. This figure of merit will be described in section 2. Then, section 3 describes the reviewed post-processing methods, section 4 develops the methodology used to compare the post-processing methods and section 5 presents the interpretation of the NQE figures.

Для продолжения чтения вы можете скачать полную версию материала по ссылке ниже