Uso de sombras na correção do efeito vinhete

Abstract

When registering spectral radiance from surface targets, digital numbers recorded by the imagery sensor may vary. Such variation causes imperfections on the images coming from aerial surveys. Variation in the image brightness related to the distance from the center of the image is known as the vignetting effect. Correcting this effect aims at achieving an homogeneous image brightness. The purpose of this paper is to present a specific methodology to determine a model in order to minimize this vignette effect based on a model fit by Least Squares Method (LSM), using digital numbers (DN) from shadowed regions. The main hypothesis is that the recorded DN of shadow pixels should be suitable to model the vignetting effect. Considering that the vignetting effect could be modeled as a trend of spatial image variation, a trend surface analysis of a sample of pixels from shadowed regions was carried out. Two approaches were adopted to represent the shadow regions of an image. The first one takes into account the components R, G, B of the aerial image within the visible spectral band, and the second one considers the component I of the HSI image. In order to evaluate the methodology, a study case with a color aerial image was carried out. The findings showed that the best results were obtained by applying the model in the RGB components, which allows to conclude that the vignetting effect can be modeled based on trend surfaces fit on shadow regions DN.