Active Pixel Sensors (APS) CMOS Image Sensors (CIS) are the most common image sensors on the market these days and part of every mobile phone and every other device containing a camera (which these days includes many surprising appliances, possibly including your toaster). They are characterized, among other properties, by extreme noise suppression performance, such that they can record charge generated by incident light with only a few electron charges imprecission, resulting in crystal clear high resolution images.
However, reading out an image from that sensor destroys/uses up the accumulated charge. So if the image is flawed, e.g. underexposed, and you (or some fast algorithm) realize this only after looking at the image, you'll have to make a new attempt to retake the image with a different exposure time or other parameter adjustment. In high speed applications (e.g. automotive applications), you might not have the time for this.
Thus this project wants to explore some ideas for 'non-destructive' read-out of the image, or just regions of interest (ROI) in the image, where the accumulated charge is preserved in the pixels (albeit with more noise than the destructive read-out). Then this preliminary image can be analyzed and a decision can be made to either stop exposure or continue for some time with adjusted parameters. A particular challange will be not to interfere with the final high quality destructive read-out, or to at least keep the loss of quality minimal.