Practical anti-vignetting methods for digital cameras

Abstract

Vignetting is a position-dependent light intensity falloff commonly found for commercial digital cameras. An anti-vignetting method is concerned with the estimation of a correction factor at each pixel position. In this paper, we propose two anti-vignetting methods which effectively estimate the distribution of correction factors. The first method utilizes wavelet-based denoising for efficient suppression of input image noise. Decimation of the smoothed distribution of correction factors is then carried out. The second method is more concerned with the appropriateness for embedded digital imaging applications. We approximate the distribution of correction factors by a single correction function, which is in the form of a 2-D hypercosine function. Only five parameters are needed to describe an underlying input intensity distribution and are estimated by nonlinear model fitting against measured input illumination data. We show the performance of the proposed methods by experimental results using synthetic and real images.