<title>Noise modeling for design and simulation of computational imaging systems</title>

Abstract

Understanding signal and noise quantities in any practical computational imaging system is critical. Knowledge of the imaging environment, optical parameters, and detector sensitivity determine the signal quantities but often noise quantities are assumed to be independent of the signal and either uniform or Gaussian additive. These simplistic noise models do not accurately model actual detectors. Accurate noise models are needed in order to design optimal systems. We describe a noise model for a modern APS CMOS detector and a number of noise sources that we will be measuring. A method for characterizing the noise sources given a set of dark images and a set of flat field images is outlined. The noise characterization data is then used to simulate dark images and flat field images. The simulated data is a very good match to the real data thus validating the model and characterization procedure.

Imaging system MTF- modeling with modulation functions

“…The major part of an imaging system consists of an input opto-electrical converter (optics and image sensor) and transmission of data (compression) for display using an electro-optical converter (Liquid Crystal Display and other monitors) [2]. Performance can be gauged in terms of contrast (degrees of gray) or modulation and is closely related to the degradation of the image produced by the imaging system.…”

Digital camera simulation

“…There has been much progress in modeling the individual components of a digital camera, including camera optics [1][2][3] and sensor noise [4][5][6][7][8][9]. Progress has also been made in modeling the spectral reflectances and illuminants in a scene [10][11][12][13][14][15][16][17].…”

Noise Measurement for Raw-Data of Digital Imaging Sensors by Automatic Segmentation of Nonuniform Targets

“…Nevertheless, to fully exploit the potential of these modern algorithms, it is necessary that an accurate noise model is used when processing the raw-data. Noise modeling for digital sensors and their raw-data has been a vivid research area in the recent years (e.g., [1]- [3], [9], and [10]). …”