Comparing 2 Images with shifting

[sand0000]

Ok, so I'm working on a project that takes a picture of a black and white checkerboard LCD and checks for dead or stuck pixels. I have something that works until the screen moves even the tiniest bit. This is on a manufacturing floor so the screen will move. How do I take into account that the screen might move a little and still keep my program sensitive enough to find a dead pixel? 

[Mark_Yedinak]

You will need some fixed points in the image that you can detect which will allow you to orient the image correctly. Not knowing you specific setup, but it you are literally taking a picture of a screen and analyzing it, you can place some colored squares on the edges of the screen (outside of your area of interest) and detect where those squares are in your image to analyze. Based on those known positions, you can then orient your actual area of interest from the image and analyze that. However, if you need to get down to the pixel level in terms of accuracy, this is still easier said than done.

[sand0000]

Ok, may have to try that. First we're trying brackets to ensure that the screen doesn't move AT ALL when we open and close the fixture. That has been the main issue so far. We're thinking a hardware solution may be simpler at this point. Sorry, can't share more detail. 

[Mark_Yedinak]

Generally when doing image processing it is a good idea to have reference points in the image so you can orient it correctly. Of course, this depends on the actual application but in your case I think they will be needed.