image zooming/enhancing software?

Concept number one: You can't manufacture detail that isn't there in the image in the first place. So most of the TV stunt-technology you see has a grain of truth but is way overblown - like extracting the licence plate of a car around the corner and down the block by enlarging and enhancing a dime-sized reflection on the chrome bumper of another car... Nah. Unless the camera was like, 300 Megapixels and the bumper was a perfect mirror. Any details that are below the pixel size of the image capturing device (pixels on a digital camera, silver grains on film) cannot be recovered.

When you enlarge an image, you get the same detail, in larger pixels. The number of pixels doesn't change, unless you use interpolation. This is using math to estimate what the intervening pixels would have been coloured if they existed. Example: You want to blow up an image to double its size. So a 1 Megapixel image will become 4 megapixels -- the other three million pixels don't exist and they have to be manufactured.

ACG The original information (letters stand for colour values)
A_C_G Now stretched out to 2x size, leaving blank spaces
AbCeG Interpolated by averaging the colour difference in the "spaces"
and inserting new pixels
(of course this has to work in two dimensions)

Interpolation works well for areas of little detail and smooth colour, it works less well for contrasty images and sharp edges. This is a simplistic explanation, the math that Photoshop uses is much more complex.

LizardTech has a fractal mathematics software Genuine Fractals that does a better job than Photoshop can on its own. http://www.lizardtech.com/products/gf/

Thanks
Trevor
CanadaRAM.com

Do any of these packages or techniques take jitter into account? If you take several nearby frames in a video recording, you can take advantage of the fact that the camera moves slightly, in some instances, to enhance resolution by combining the images. In fact any interlaced display essentially creates sub-pixel rendering by essentially the same technique, employed in reverse.

But yeah, for the most part, this is pretty fugazi on TV.

EDIT: Trevor, have you used that Genuine Fractals product? Is it really substantially better than PS's Unsharpen Mask? I wasn't able to find any really informative samples on their website. Even the one in the 5MB (!) datasheet wasn't really set up in a way that was helpful. 🙁 It also doesn't say if the plugin works in PSE. I guess I should download the trial and find out, though! 🙄

mkrishnan said:

Do any of these packages or techniques take jitter into account?

Click to expand...

Strange, I was thinking about just that as I was writing it up. If you took samples in rapid succession with some way of knowing exactly what the motion/angular change was, you could calculate sub-pixel details from how the patterns "migrated" from one pixel to another. You would either need to have phenomenal lockup with the motion, or you would have to derive motion from common elements in the scene and calculate it backwards.

Reminds me of a camera someone was developing with a spinning lens or something, which could capture an image with infinite depth of field. They were stitching together the "in focus" portions of many successive (or actually continuous) exposures. Looked like %#*$ though...

mkrishnan said:

EDIT: Trevor, have you used that Genuine Fractals product? Is it really substantially better than PS's Unsharpen Mask? I wasn't able to find any really informative samples on their website. Even the one in the 5MB (!) datasheet wasn't really set up in a way that was helpful. 🙁 It also doesn't say if the plugin works in PSE. I guess I should download the trial and find out, though! 🙄

Click to expand...

Yes, I have it -- it's actually made me money because I have sold photos to magazines that I "up-megged" from the original jpg files.

I think it does a better job than PS 8, anyway.

But it's not the Unsharp mask you're comparing it to - it's the bicubic interpolation when you change Image Size.

One thing you can do with either program is to increase the image resolution about 4 x through interpolation (some say it works better in PS if you do it in successive small jumps rather than one big one). You are going to get some noise in the image, so Unsharp Masking now would just accentuate the noise. So while it is at the largest size, apply Dust and Scratches or Despeckle to smooth out the noise (you'll blur some detail, too). Use selections to apply more blur or despeckle to background only, if you wish, preserving the detail in the foreground.

Then reduce the size again to the target size. This will gain you back the detail that you lost through despeckling, because the 4-pixel-blurred edges will come back down to single pixels.

Now add your unsharp mask to enhance detail without enhancing the noise.

You should only ever apply USM at the end of the process, and at the final resolution. Scaling USM'd images gets ugly.

CanadaRAM said:

Yes, I have it -- it's actually made me money because I have sold photos to magazines that I "up-megged" from the original jpg files.

Click to expand...

Nice! 😎

I am going to give it a try...although I have a hard time seeing myself paying $160 for a sharpening plugin, as a hobby artist.... 😱