Camera Response Function Estimation from a Single-channel Image Using Differential Invariants
Tian-Tsong Ng, Shih‐Fu Chang, Mao‐Pei Tsui · 2006
The camera response function (CRF) models the important characteristics of cameras and digital images. Recent work in [1] proposed a method for estimating CRF from a single grayscale image, by assuming the uniform distribution of the image irradiance in local edge regions. In this paper, we propose a different approach by exploring the fundamental properties of digital images ‐ differential invariants. Specifically, the geometry invariants can be used to derive the analytical and computational solutions for estimating the CRF, based on the detailed analysis of the invariants. As the technique needs an analytic CRF model with good modeling power, we propose a general polynomial exponent model. The proposed method is tested using images from four different models of digital camera and achieves a very good accuracy ‐ the average root mean square error (RMSE) for the estimated CRF is 0.0369. Being the basic properties of digital images, the differential invariants also show a great potential in solving other problems, such as modeling the cameras with a spatially varying CRF. This is a current ongoing work.