Nice PDF. Some cool new stuff, smart decisions & remarks (noise, hotpixels, matching frames, etc).<p>If you have an iOS device, check out the app Average Cam Pro[1]. I've had it on my iPad for several years now. It does an awesome job taking noise-free, detailed pictures (as long as you don't hold it in your hand) by taking up to 100 pictures and merging them. You can adjust exposure etc. after the fact. Great stuff.<p>I have yet to find a good replacement for my Android phone. The closest thing to [1] is Multiple Exposure Camera [2] which has a terrible interface and no exposure setting, but some other nifty settings to remove moving objects etc. (note: just don't use the 0s interval or it crashes)<p>[1] <a href="https://itunes.apple.com/us/app/average-camera-pro/id415577873?mt=8" rel="nofollow">https://itunes.apple.com/us/app/average-camera-pro/id4155778...</a><p>[2] <a href="https://play.google.com/store/apps/details?id=com.conslazy.multipleexposurecamera&hl=en" rel="nofollow">https://play.google.com/store/apps/details?id=com.conslazy.m...</a>
That was a great read, back in the early 80's at the Image Processing Lab at USC there was a project to turn, what was then, a new "CCD" imager into something that could collect images that were useful for image analysis. At the time these "TV" cameras were equivalent to a really low end web camera. What we came up with was basically the first half of this talk, taking frame after frame and integrating it into the frame buffer with a filter function. Sad that I can't find the technical reports in Google Books.
Very impressive! I have an app on my android that follows the same basic idea (accumulate frames) but doesn't come anywhere near this.<p>Unfortunately the author, in a reply to a youtube comment, stated that this is work he did for Google and doesn't know if the company has any plans for this.
There is an old free software package ALE ( <a href="https://unix4lyfe.org/ale/" rel="nofollow">https://unix4lyfe.org/ale/</a> ) that does some of the processing like this. It was much easier to get working if you had linear light images... and registration of really noisy images was quite hard on it.
Here's a (v short) video by the paper's author showing the app in action with there manuscript that features early on in the paper: <a href="https://youtu.be/S7lbnMd56Ys" rel="nofollow">https://youtu.be/S7lbnMd56Ys</a>
In php is quite simple to merge images by pixel,
I also tried to delete moving object
<a href="https://github.com/br1n0/phpPhotosMerger" rel="nofollow">https://github.com/br1n0/phpPhotosMerger</a>
Is this what makes the google pixel phone camera so outstanding? it would explain why a technically inferior camera (slower aperture, no optical image stabilisation) outperforms both apple and Samsung flagship phones.