Computers detecting inferior printing quality and errors

As observed, the described approach to computer-aided error detection in printed matter proves itself to be exact and reliable. Our prototype printing error detection computer program has also undergone extreme testing in an industrial environment. The problem of excessive detection duration will have to be solved by an optimization of the registration and interpolation procedures. When considering the coarse image registration, a solution in the frequency domain seems appealing, where the translational and rotational information can be extracted from the phase and amplitude spectra, respectively. Fine image registration must be optimized even more. One of the ways we are presently researching tries to minimize the quantity of data included in the computation, at each iteration. We can do this by using images with lower resolution or taking into registration only an image's segment. Afterwards, the obtained registration parameters would be applied to the original-size images. For the time being, a second optimization attempt with partial images seems the most promising.