Empirical Conditional Mean: Nonparametric Estimator for Comparametric Exposure Compensation

In this paper, a comparametric exposure compensation is conducted using a nonparametric estimator: empirical conditional mean. The Nadaraya-Watson estimator is used to smooth the empirical conditional mean curve especially for the case of small number of samples. The performance of the estimator is compared with those of the polynomial and piecewise-linear fittings. Designing the Nadaraya-Watson estimator is very simple and achieves lower errors than the fitting cases, which require a heavy computational burden of solving equations, without worry about the singular matrix case