Vignetting Correction Using an Optical Model and Constant Chromaticity Prior

Beinan Yu, Jiacheng Ying, Lun Luo, Si-Yuan Cao, Xiansong Bao, Hui-Liang Shen

Abstract: Vignetting correction is a key pre-processing module for most imaging systems related to computer vision applications. Vignetting estimation is difficult for the current imaging systems due to the complex optical components and challenging vignetting feature extraction. In this work, we propose an algorithm for single-image vignetting correction. We present an optical aperture limit model that uses an occlusion parameter to estimate vignetting effect caused by light occlusion. Moreover, we derive a novel prior that the pixel chromaticity is not affected by vignetting. Our algorithm efficiently selects regions for feature extraction according to the chromatic and spatial analysis of the image, and then uses the intensity ratios to predict the vignetting model parameters and optical center guided by the constant chromaticity prior. We evaluate our algorithm on both synthetic and real-world images. Experimental results indicate that, compared with the state-of-the-art methods, our algorithm achieves better correction under different types of vignetting.