Sensor Noise Modeling by Stacking Pseudo-Periodic Grid Images Affected by Vibrations
Abstract
This letter addresses the problem of noise estimation in raw images from digital sensors. Assuming that a series of images of a static scene are available, a possibility is to characterize the noise at a given pixel by considering the random fluctuations of the gray level across the images. However, mechanical vibrations, even tiny ones, affect the experimental setup, making this approach ineffective. The contribution of this letter is twofold. It is shown that noise estimation in the presence of vibrations is actually biased. Focusing on images of a pseudo-periodic grid, an algorithm to discard their effect is also given. An application to the generalized Anscombe transform is discussed.
Origin : Files produced by the author(s)
Loading...