Hdr Image Reconstruction using Locally Weighted Linear Regression

摘要

In this paper, we propose an HDR image generation method from a single low dynamic range (LDR) image based on its inverse camera response function (ICRF). In the method, the ICRF is modeled and estimated using Bayesian framework firstly; then the details in over-exposure region of the input LDR image are enhanced; finally, the high dynamic range (HDR) image is generated by conducting the ICRF on single LDR image using locally weighted linear regression (LWLR) calibration. We demonstrate our method on images acquired with both photochemical and digital imaging processes. Extensive experimentations show that accurate calibration results can be obtained using only a single input image, and the approach induces less contrast distortion.