Improved calibration model for single-camera endoscopic tomographic systems

Multidimensional imaging has become one of the major developing trends of combustion diagnostics, and volumetric tomography is one of those techniques that has experienced significant progress over the past decades. Numerous time-resolved modalities of volumetric tomography have been developed to image a variety of physical quantities. Due to formidable expenses associated with the high-speed cameras, single-camera endoscopic versions have become more and more popular. Calibration is a critical step to establish the geometric relationship between the projections and the volume of interest. Zhang’s calibration model [ Seventh IEEE International Conference on Computer Vision (ICCV) ) (IEEE IEEE , 1999 ), Vol. 661 , pp. 666 –673 ] for a multicamera system has been adopted extensively for endoscopic tomographic systems. However, Zhang’s model is insufficient to depict the entire imaging process due to the existence of fiber bundles. In this work, an improved calibration model is developed by introducing a secondary imaging process to account for the introduction of the fiber bundles. Comparative studies were then conducted both numerically and experimentally to assess the imaging models. The results showed that Zhang’s model could introduce a large error in distance estimation when the magnification is nonunity, while the modified model can achieve a higher calibration precision under various secondary magnifications. This work can help further improve the reconstruction accuracy of endoscopic tomography.