High res image from video

[Mikezzz]

I would like some software that could produce a high resolution image from a video clip or possibly from just a collection of still images of the same subject.

Thanks
Mike

[Ritho]

Are you saying you want to take a number of frames from video and piece them together in to one larger picture? First you need to capture each frame you want to use and save them as individual images. Then you can use any one of these http://www.techsupportalert.com/best...e-stitcher.htm But Microsoft ICE will do exactly what you want amazingly well. All you have to do is put all the images in a folder, and tell ICE to use them all. It figures out where each image belongs in the scene and "mashes" them all together i without you doing much of anything.

If you want to take it up a notch, you can take all your pictures from one perspective, walk around to another place and take a bunch of pictures of the same scene, then do it again, and again until you have made a 360 degree circle of the scene you want. Then using the big brother of ICE called Photosynth you can make them into a 3D image that you can move around in and see from every perspective you took the images from. Very much like Google Maps Street View if you have used that. http://photosynth.net/about.aspx

[Mikezzz]

Thanks for the feedback. I have used ICE before. I just tried Photosynth. I was very impressed.

Neither of these do what I want though.

I am not wanting to create a panorama or 3d space etc.

I just want to extract a single image from a video. Since the video will have lots of frames very similar, I am hoping that some software will be able to get some of the data from adjacent frames to improve the image quality, possibly increasing the resolution of a single frame grab.

Thanks
Mike

[Ritho]

Mike, Vreveal might help with what you want to do http://www.vreveal.com/download

What I think you want to look for is Video Forensic tools. That is the way I found Vreveal. There are methods of interpolating the data from a group of video frames to produce a better quality image, but I don't know if you could really call the resulting image "high-resolution" but it would have more detail.

Edit: In fact it is call Super Resolution. See my next post.

[Ritho]

The technique is called Super-resolution http://en.wikipedia.org/wiki/Super-resolution

Here is a link to a technical document on the technique. www [DOT] cs [DOT] rhul.ac.uk/home/fionn/papers/fm7 [DOT] pdf

You might be able to use this software in some way. http://www.astronomie.be/registax/

[crank]

Averaging out multiple frames is a great way to reduce noise, but you really can't ever increase the resolution. I always moan when I see CSI or other shows/movies where they take a really lousy security camera image and by the magic of computers read a license plate number where originally you barely could tell there was even a license plate there.

If the information isn't there in the original, processing can't extract it [it can create it though, as our brains constantly do and we are oblivious, only what is created may have zero to do with reality]

Super Resolution from what I just gleaned is able to perform some near magic, but requires quite restrictive conditions/foreknowledge etc. Think of trying to listen to a song on an extremely staticy radio, first you have no idea what it is, but if you manage to hear enough to figure it out, suddenly you can hear the music quite well. Your sound processors are using their knowledge of the signal to assist in filtering out all the noise. The radio station didn't suddenly start coming in better.

Avidemux is a great program for extracting the stills, it can easily save all the frames from a selection to the whole video.