In some papers, this problem is partially solved by estimating GT by experts in the field (biologists or physicians). However, in many cases such GT estimate is very subjective and strongly varies among different experts.
In order to overcome these difficulties we have created a toolbox that can generate 3D models of artificial biological objects along with their corresponding images degraded by specific optics and electronics. Image analysis methods can then be applied to such simulated image data and their results (such as segmentation or measurement results) can be compared with GT derived from input models of objects (or measurements on them). In this way, image analysis methods can be compared to each other and their quality (based on difference from GT) can be computed.
The present version of the simulation toolbox can generate cells in 3D using deformation of simple shapes and adding texture to the cell interior. Further, it can simulate optical degradations using convolution with supplied point spread function as well as CCD camera artifacts such as impulse hot pixel noise, additive readout-noise or Poisson photon-shot noise.
We have also dealt with the task of evaluating the quality of the simulated images in terms of their similarity to real image data. We have tested several similarity criteria such as visual comparison, intensity histograms, central moments or entropy.
The talk will provide a short overview of 3D microscope image formation, mention specifics and problems of fluorescence mode, present examples of applications and finally describe the simulation process and evaluation of the quality of simulated images.