Sharpen detail by averaging successive frames

RickD wrote on 4/2/2016, 11:47 AM

I have some old videos from which I am trying to extract background detail. A typical scene is a head shot of someone talking with signs on stores, street signs, building addresses, etc. visible in the background. The camera stays in place for at least 15 seconds.

Can anyone recommend a method in Vegas for sharpening this background detail? I am thinking it could possibly be done by averaging a number of frames. The end result should be a single image with sharper detail in the background. Obviously, averaging frames would tend to blur the person speaking but that's OK because I am only interested in the stationary background.

Comments

Chienworks wrote on 4/2/2016, 12:19 PM

This is what PLAY technologies used to do with their early capture device that claimed to pull 5MP images out of a VHS capture. As far as doing it in Vegas, stack up as many individual frames on separate tracks, lined up vertically, then reduce the opacity of all but the bottom track. How much? Well, if you have two tracks and you want equal parts of each the top should be 50% and the bottom 100%. For three tracks use 33%, 50%, and 100%. Four tracks: 25%, 33%, 50%, 100%, and so on. Each additional track would be set for 1/(number of tracks) opacity to get an even blend.

RickD wrote on 4/2/2016, 1:12 PM

Thanks, I tried your suggestion but did not achieve the results I wanted. It did blur the person talking as I assumed it would but the background did not get clearer. I assume that my theory about sharpening the motionless background using this method was not correct.

What if I were to just ask the question how to sharpen the background without being concerned about foreground action, and without offering an initial suggestion? Any ideas?

wwaag wrote on 4/2/2016, 3:40 PM

One option would be to duplicate the track and create a feathered mask around the talking head. Then apply sharpening only to the track where the background is being shown. This way, the talking head would remain unchanged while the background is sharpened. Obviously, the usefulness of this approach would greatly depend on how easy it is to create a good mask.

wwaag

wwjd wrote on 4/2/2016, 3:48 PM

SHARP And UNSHARP MASKa higher copy track, and COOKIE CUTTER a hole for the foreground guy?

PeterDuke wrote on 4/2/2016, 6:06 PM

If the camera and background are truly stationary, then each frame of the background will be identical, so averaging will make no change. Any random noise should be reduced however.

I have an old Win XP program called Red Hawk Paparazzi that works on slowly moving material, such as from a hand held camera. It works with DV AVI interlaced footage and re-aligns each field for best match. You could average up to about 6 fields, I think. It sometimes gave a modest improvement on a single frame. but it was never HD quality.

larry-peter wrote on 4/3/2016, 11:39 AM

There is an image enhancement technology that will do what you're looking for - and as with all enhancement techniques, it has limitations. Google "super resolution image enhancement" and you'll find examples of what can be accomplished and its limitations. This appears to be not a simple blending of frames, but adding only "new" data from surrounding frames to the base frame to increase detail not present in a single frame. I don't think you're going to get what you're looking for by staying within Vegas.

riredale wrote on 4/4/2016, 9:50 AM

There is a program used in amateur astronomy called Astrostack. It allows one to stack multiple stills or frames in order to greatly reduce noise and/or the effects of the atmosphere.

It can NOT create detail finer than what the optical system can theoretically deliver. But take a look at the home page to see the remarkable improvements it can do.

Serena Steuart wrote on 4/5/2016, 9:01 PM

I prefer http://www.astronomie.be/registax/Registax[/link]. It has excellent control of sharpening.