Direct Image Registration With Gain and Bias

Image registration consists in estimating geometric and photometric transformations that align a template and an image as best as possible. The direct approach consists in minimizing the intensity discrepancy between the aligned template and image. The inverse compositional algorithm has been recently proposed for the direct estimation of groupwise geometric transformations. It is efficient in that it performs most computationally expensive calculations at the pre-computation phase. We propose the gain and bias inverse compositional algorithm which estimates, along with the geometric transformation, a photometric one modeling for example global lighting change. Our algorithm preserves the efficient precomputation-based design of the original inverse compositional one. Previous attempts at incorporating appearance variations to the inverse compositional algorithm spoils this property. We report experimental results on simulated and real data, showing the improvement in computational efficiency of our algorithm compared to previous ones.