locked
ICE image resolution enhancement ? RRS feed

  • Question

  • Hi µsoft research dev team !

    I just tried out the ICE image composition soft from your labs to convert a video to a panoramic picture.

    1 word comes to my mind:

    AWESOMENESS !!!!!

    Though i'm not one of the best supporters of µsoft products and spirit, i gotta admitt that this lil software is nothing but

    honey for those who do some graphic thingies. And if, as it seems to be the case with ICE and other tools like live video, µsoft

    comes an walk on linux places ( the world of free software ), i must confess that my old regrets from this free world will quickly

    fly away ! and i will surely revise my opinion about your multibillion$ company !

    But please lemme talk about what i'm here for:

    ICE is great ! but...

    It would be even greater if...

    pictures resolution would be enhanced. lemme explain:

    A video is a following of 'almost identical' images. When you do a panoramic with your camcorder, you often can see on successive images of the video, the same object. those successive images could be used ( through some lil image processing soft ) for enhancing the final image resolution, with some kind of transfert func like: final pixel with final res = sum of means of surrounding pixels through a weighted matrix. This is some kind of image resampling using time and space ( time = frame time, and space = 2D space reconstructed from image movement ) as parameters. Imagine a flower in a grass land. You do a left to right pan on this land. The flower appears on the right side of the pic, stays for say... 80 frames, and disappears on the left side of the pic. during thos 80 frames ( specifically talking about this flower ) the flower pixels can be located on the picture, and define more pixels on the final picture than there was on entry video picture. This priciple is used i think on space images enhancements for sky photos. With 20 or 30 photos of the same place in sky, you can get much more detailed images than with one, or even with CCDs 20 or 30x the base resolution.

    I admit this could lead for ICE to huge final pics.... but today, as the base comps all have at least 4Gb of ram, this problem is pushed away, and it would be the choice of the user to have 4, 9, 16 or even more times detailed final pictures. ICE would provide hi-res images, and user can do whatever he wants with it.

    This might be a lil idea. Maybe not quite usefull for all common uses, but, as ICE already does image processing with chunk relocation and motion estimation i thing this wouldn't be quite hard to implement resolution enhancement in it. Imagine ! You could do what hollywood films show off !!!! Extract a man face from a 4 pixels photo !!!!! But this time, for real !!!! ;-D

    Finally i must congratz all the team for this good work ! and also for this kind of mind that i didn't know in µsoft:

    share with the world ! The world with share with you.

    best regards.

    PS: Sorry for my approx english. I'm french and i confess i don't master it correctly...




    • Edited by brefromjeu Friday, August 17, 2012 12:58 PM
    Friday, August 17, 2012 12:50 PM