Self-Calibration of Fused Camera Images

Abstract. The standard sensor for large-format digital cameras is currently the Sony CMOS chip with approximately 14200 × 10600 pixels of 3.76 μm. Sony CMOS chips with ~ 19299 × 12800 pixels and a pixel size of 2.81 μm will be available in near future. The very large sensors used in the DMC-3 are no longer manufactured. To achieve larger imaging systems, camera systems with multiple sub-cameras are being used instead of single cameras. The images of the sub-cameras must be fused into homogenous images. The dominant method for the image fusion is the geometric fusion of the sub-images to lower resolution overview images. To achieve this, the images are geometrical enhanced by the calibrating the sub-cameras. Theoretically, the fused, geometrical enhanced images should be free of systematic image errors. However, this must not be the case, even with thermal control of the sub-cameras, satisfying thermal camera control of the entire camera system is not possible. The standard additional parameters cannot be used for image fusion problems that do not meet the calibration of individual cameras. Therefore, a specific set of additional parameters is required for each type of fused images. It turns out that full geometric accuracy can only be achieved with such special set of additional parameters, but it is possible to reach the accuracy as for the single cameras.