I have the onOne Perfect Resize 7 Pro, which I really like, but I have a question about down-sizing images.  Generally, Perfect Resize is used to increase the size of an image, but does anyone know how it stacks up against Photoshop or other apps./filters as far as decreasing an image size?  I haven't had much luck finding information on this topic.

I haven't used the program One Perfect Resize. I have used a freebee that, IMHO, is incredible. Irfran View can down size an image in just about any way you can think of. Porpotional, percentage, length, width, preserve aspect ratio and a whole bunch more. It  works with a kazillion different formats.

Attached Link: downsampling tests

Theres no guess work or 'invention of data' involved. All the available algorithms even simple 'Nearest Neighbour' can give quite good results.

That doesn't mean there aren't better methods for specific occasions.

Counter intuitive ones like pre-blurring before reduction for example.

Have a look at the above page for some old experiments on downsampling methods.

edited to correct the spelling of neighbour! :biggrin:

Quote - I think the main reason theres not much information on reduction is because its relatively straight forward compared to image expansion.

Theres no guess work or 'invention of data' involved. All the available algorithms even simple 'Nearest Neighbour' can give quite good results.

 

 

edited to correct the spelling of neighbour! :biggrin:

Hmm...... I'll have to think about that a bit, but intuition tells me an image, whether scaled up or down, requires "decisions" of what color to apply, remove, or change for new/neighboring pixels.

In Photoshop, the same algorithms new or can be applied scaling up or down.  Perfect Resize uses a different algorithm to scale up than PS does, so I was assuming it would (use that algorithm) scaling down.

Upscaling and downscaling are very different beasts.

The empty blank areas in an upscaled image have to be filled with newly created pixels. Thats what I meant by 'invented data'. They are 'guess-timated', there is no mathematical way of knowing they are correct. All sorts of techniques, algorithms and recipes have been invented for this purpose. Specialist ones are better then anything built into photoshop.

With downscaling there is no unknown area to be filled in. With all pixels known it becomes more of an averaging problem.

I'd be very surprised if one size Pro used the same algorithm for both.

The different algorithms have specific applications in PS too. For example Bicubic Smoother is for upscaling and Bicubic Sharper for downscaling.

..... And that's why I asked the question to begin with, becuase I can see quite a difference between the various downscaling algorithms and was wondering how Genuine Fractals (Perfect Resize) stacks up.

Never heard the question asked...

Yeah, generally no one puts much thought into downsizing algorithms because they've been done again and again by so many industries (not just imaging, think video as well), and the logic is all the same - you're trying to take OUT details because you're removing pixels/detail.

You reduce something /5 and you have to figure out which 4 pixels are to be thrown out so you can only have 1 pixel left - what is the dominant color for it to be to "simulate" what detail is now gone - get it?

Yes - the big deal has always been UP-sizing since pixels need to be manufactured that were never there in the first place. How do you keep clean lines instead of simply bigger jaggies? How do you deal with the notion that the subtle colors were actually a cloth texture pattern? Stuff like that?

Just about ANY program will do the trick well when it comes down to down-sizing - any differences they may end up having will simply be so minimal no one would bother caring.

Hope this helps-

-Lew