Issue No. 04 - April (2008 vol. 30)
Seon Joo Kim , IEEE
Marc Pollefeys , IEEE
In many computer vision systems, it is assumedthat the image brightness of a point directly reflects the sceneradiance of the point. However, the assumption does not hold inmost cases due to nonlinear camera response function, exposurechanges, and vignetting. The effects of these factors are mostvisible in image mosaics and textures of 3D models wherecolors look inconsistent and notable boundaries exist. In thispaper, we propose a full radiometric calibration algorithm thatincludes robust estimation of the radiometric response function,exposures, and vignetting. By decoupling the effect of vignettingfrom the response function estimation, we approach each processin a manner that is robust to noise and outliers. We verifyour algorithm with both synthetic and real data which showssignificant improvement compared to existing methods. We applyour estimation results to radiometrically align images for seamlessmosaics and 3D model textures. We also use our methodto create high dynamic range (HDR) mosaics which are morerepresentative of the scene than normal mosaics.
Radiometric response function, vignetting, radiometricimage alignment, high dynamic range imaging
S. J. Kim and M. Pollefeys, "Robust Radiometric Calibration and Vignetting Correction," in IEEE Transactions on Pattern Analysis & Machine Intelligence, vol. 30, no. , pp. 562-576, 2007.