Errors in imaging systems

Both image capture and image reproduction inevitably result in imaging errors caused by the non-ideal real-world characteristics of the transmission chain.
The illustration shows the problems that occur using the example of a regular stripe pattern recorded with a CCD line scan camera. Ideally, the image signal would be a regular rectangular curve. In practice, however, the signal deviates significantly from this.

The following typical image errors occur:

• Blurring
• Geometric distortions
• Uneven illumination
• Color divergences

To make matters worse, these errors do not occur evenly throughout the image, but are more pronounced in some areas and less so in others. These are referred to as spatially variable image errors. It is precisely this spatial variance that complicates the application of classic image restoration methods such as inverse or Wiener filtering. The present image correction method therefore uses an artificial neural network that takes all these types of imaging errors into account simultaneously and spatially.