Scenario: Stationary view with sequential timestamped samples in 2D-color where the comparison is only done between one pair of neighboring samples.
- Have never used OpenCV before,
- Limited understanding of computer vision,
- Moderate experience scripting programs.
- Given the information provided, what additional information would be of use?
- What existing related analysis and reporting solutions within OpenCV should I look into more?
- What is the source of these images? I've abstracted the intent of a problem to the degree in an attempt to avoid skewing how the answers conceptualize the related solutions. Disclosure-of-Problem-Domain: It's possible that I'm wrong about the value of abstracting the problem domain, and since it's already been guessed, I'll confirm the domain is any high-value automated GUI testing that is of relatively low-complexity. Currently, seems like simple regression testing based on pixel-for-pixel change which is then sorted by the percentage of pixels change on a pair-by-pair basis would be quick and produce value, but the template matching suggested by Martin Beckett might allow for cross-platform testing, though again not knowing much about OpenCV, or the domain of computer vision, it's hard to say complex or valuable such an approach might be currently.
- What is the same? Depends on the solutions OpenCV offers in the context of image difference analysis and reporting. Any artifacts related to the capture would/should be accounted for in the comparison analysis. Ideally, difference reporting should by default (or predictive coding) only profile material significant state changes in a way that allows the manual response to them to be prioritized.
- What are the circumstances of the image capture? All pairs are sampled within a highly controlled environment with the same stationary view/angle, lighting, exposures, focus, etc.