Improved Weighted Least Square Radiometric Calibration Based Noise and Outlier Rejection by Adjacent Comparagraph and Brightness Transfer Function

摘要

In this paper, we propose an improved radiometric calibration algorithm by extending the previous radiometric calibration method that is based on the "fixed exposure ratio" Brightness Transfer Function (BTF) and Weight Least Square. Our proposed method can be applied to the radiometric calibration on the case that the images are captured with non-constant exposure ratios which makes the image acquisition process to be more flexible. The key idea of our new proposed method is to determine a set of BTFs between the set of consecutive image pairs and then to form the set of composited functions. From the composition, we then convert the BTFs of all image pairs into the virtually single BTF. The noise & outlier rejection based on this BTF is applied to retrieve the clean data. Finally, a reformulation of weighted least square minimization is proposed to estimate the camera response function (CRF) of a camera. From the performance of our proposed algorithm in comparison with the state-of-the-art least square method proposed by Mitsunaga and Nayar as the baseline, we found that our method outperforms the baseline algorithm on both the synthetic dataset and the dataset of real-world images.