search

junyanz / light-field-video

Light field video applications (e.g. video refocusing, focus tracking, changing aperture and view)
http://cseweb.ucsd.edu/~viscomp/projects/LF/papers/SIG17/lfv/
203 stars 44 forks source link

Rehexing Images #4

Open crvogt opened 6 years ago

crvogt commented 6 years ago

Hi there, I was just wondering, seeing as the output is 5xx by 3xx pixels, why don't you decompose the sub aperture images back into the raw format, and synthesize your image that way (ie, the way the lytro software might)? It seems like the output could be much higher resolution.

junyanz commented 6 years ago

@tcwang0509

junyanz commented 6 years ago

I didn't fully understand your question. Could you elaborate on that?

tcwang0509 commented 6 years ago

You can of course reshape all the viewpoints into one image, but the spatial resolution still remains the same. You won't get extra pixels by just doing that.

crvogt commented 6 years ago

Thank you both for answering, I'll of course attempt now to make my question clearer :)

I assume that you sampled the individual micro lenses to get the resulting sub aperture images (similar to LFToolbox if you're familiar), unless lytro does this for you, ie pulling one pixel from each of the lenses gives you one sub aperture image of the resolution 5xx by 3xx (depending on the number of micro lenses).

The Illum has a 40MP(?) sensor, and I assume using their rendering algorithm they get images of at least 4MP when displaying the pictures you've taken with the camera in their software suite, so why not reform your sub aperture images back into the raw micro lens image style (one pixel from each viewpoint going to one micro lens), and render the image similar to how lytro would?

crvogt commented 6 years ago

Sorry, did that clear it up at all?

tcwang0509 commented 6 years ago

Lytro just applies some superres algorithm to the image (which I don't think is open sourced). You can always apply any superres method to increase the image resolution.