My work in Photoshop is hampered by the following obsessive suspicion. Please either bust or confirm the myth.

I believe:
Resizing a rasterized layer multiple times in a PSD will destroy its image quality.

Is it true that upon each use of the Free Transform tool, information is interpolated and/or discarded, such that multiple applications will hurt quality? I believe this and attempt to resize raster images to their destination size only once, at great time expense.

Or does the PSD somehow retain the full information of the original image, simply recalculating its new size upon each application of the Free Transform tool (or other resizer)? If that's true, my obsession would be lifted and I could work at a reasonable pace.

What do you think? Myth or fact?

Create an image file. Save it as an uncompressed TIFF. Load the file in Photoshop. Resize it. Save it as a PSD. Close Photoshop. Open the PSD in Photoshop. Resize it back to its original size. Save it as an uncompressed TIFF. Compare the two TIFF files.

What Bitt describes will alter the image with no chance of ever getting back what you had originally. There's no question that quitting the program will destroy any cached image data that might be used for resampling while the program is running.

What Mark describes is Image Resize which is not the same thing as Image Transform with manual resize. They don't even use the same resampling algorithms within the application.

It's conceivable that an application like Photoshop could cache an original layer or element such that when you resize it, the resampling is done from the information in that cache. However, I doubt very much this is the case - and it's easy to prove. Photoshop fails miserably at many basic functions/levels that I simply can't believe the engineers would ever have the desire, let alone the engineering know-how to implement something that worked properly that way. Seriously, Photoshop is a pretty basic application that relies on so much garbage legacy code that it would make your head spin. Don't get "basic" confused with "simple" - it's definitely convoluted and crufted up enough to be anything but simple.

Take a simple image that you've created yourself, preferably not continuous tone (not a photo). A crop of a screen-capture showing part of a desktop window would be good, maybe even with some text. Grab about 100x100 pixels. Paste it into a blank Photoshop document larger than the image (like 400x400) with a transparent background. Paste another copy so now you have two. Move one of the copies up to one of the corners for safe keeping. Select the other image and press CTRL/CMD-T for transform. Look at the Info palette to confirm the dimensions. Hold down shift to maintain aspect as you resize and change the scaling to some oddball value like 142% or 191%. Press RETURN to apply the change. Press CTRL/CMD-T again for transform. Again hold shift as you resize and look at the info palette so you can scale the image back to its original pixel dimensions. Press RETURN to apply changes.

Now look at the two images. They're different.

Always keep source images safe in alternate files.

Also, if you're working on anything but photos, use Illustrator instead of Photoshop in practically every case. It does a better job of rasterizing and usually a better job with resampling and using the Save For Web feature (changing color depth, compression, color remapping, etc..) In fact I even use Illustrator for isolating and resizing product photos. I will import/place a PSD at full resolution into an Illustrator document, making sure it's tagged as 72dpi to make sure no resizing is done. Then I will crate an outline to clip it from its background - the path tools/UI in Photoshop are practically useless. Then I'll use Save for Web on the clipped image to save off scaled versions with transparent backgrounds.

EDIT (added):
In fact, Illustrator doesn't do a final rasterization until you save out to a bitmap format. So you're free to resize a placed bitmap all you want within the document itself because it's only showing you a placeholder. A Photoshop-like application could definitely be designed to work this way, allowing you to save files in a composite format that would simply store each pasted or drawn bitmap object/layer at its original pixel dimensions while rendering to screen at the re-targeted dimensions. Exporting to some other non-native format would then resample and blend all the objects/layers according to your preferences and design/layout.

ImageMagick seems like not much more than a "free" implementation of many of Photoshop's features. While doing transformations in Photoshop it will continue to use the original until you apply the transformation as well. It doesn't keep resampling from the intermediate stage. That's why in the example I posted I specifically said to press RETURN to apply the transformation.

Pixelmator was just released today for Mac OS and uses ImageMagick as part of its core. It looks like a slicked up clone of Photoshop that's missing a number of features. They're certainly providing a useful tool because not everyone wants Photoshop. It's a great alternative at a much lower cost of $59. Unfortunately it falls into many of the same pitfalls. While it's using updated rendering technologies (Core Image and OpenGL) for speed, the fundamental operations are still handled the same rudimentary and archaic ways. Not terribly friendly to someone who doesn't want to use Photoshop because of its learning curve and unfriendliness.

This was a huge opportunity to make something truly new and fresh and the developers completely missed it. But they'll still make tons of money because there will be droves of lemmings drinking their Koolaid. And good for them of course.

99% of commercial software is just "good enough" - that's not the same thing as "good." Sometimes it's down right shit.

Even Lightroom with its new non-destructive editing is not very inspired. They took Apple's lead and just changed things up a bit. Though not as polished, they did implement a better asset management UI - it's also much faster on modest hardware for managing and sorting. But its retouching and correction tools are ultra weak and simply follow the same pitfalls long established by Photoshop. To put it in laymen's terms, it's just completely awkward bullshit.

Now Lightzone has a sweet set of tools. Just too many other issues with the product overall and the fact it can't replace Lightroom's Digital Asset Management nor community support.

There's still a huge hole in the market for the ideal all-in-one (or most-in-one) workflow application for digital photography. There are a lot of opportunities for more general purpose image editors as well, but this product class is a lot cloudier and easier to get lost in.

Adobe applications are the only choice around for anyone who cares about access to the same applications on multiple platforms. Sad state of affairs, because after CS3 it's obvious Adobe will never update any of their applications with the changes they need to stop being complete piles of shit. So it's shit most of us must continue to use... At least I don't have to use them under Windows.

Bruno and canuckInOR (wait, wasn't that LA?) better described my vague idea. So, the answer is:

Fact - Photoshop applies changes to the current version of the layer, not some cached copy of the original.

So, I'll just have to continue the two step process of roughly laying out my compositions, then resizing source images only once for the final production.

I mean, I only use Photoshop (7) for glorified cut and paste. I really hoped it would make that seemingly simple job a little easier.

I think a vector drawing program like inkscape will let you scale and move around multiple pasted images without losing the original resolution:

http://inkscape.org/

Ah, ok. In Photoshop, I was resampling/resizing all images on the page to the same dpi, proper size, etc, etc. That what I get for only knowing "glorified cut and paste"

I really did like working with Illustrator (5? in 2001). Sometimes I just want to draw a rectangle; do you know how hard that is in Photoshop?! I should probably pick up a copy of Illustrator.

The program i mentioned, inkscape is a lot like those other vector programs, only a bit simpler, and opensource. I think it can even open .ai files.

Two features in Illustrator that are invaluable for web graphics are its Smart Guides and Save for Web feature. Smart Guides lets you line things up without having to create guides, simply by hovering over an anchor/node of an existing object your pointer will snap to various imaginary guides coming off that point. Save for Web allows you fine control over color reduction/remapping and a multi-up preview of the original versus the intended destination formats. You can save as PNG, JPG or GIF with this feature with anything from 2 colors to 24bit plus transparency.

How well does Inkscape rasterize and what kinds of controls does it give you for saving out PNG, and JPG? I read it feature list last night, which sounds impressive, along with the FAQ and a collection of Inkscape for Adobe Illustrator Users notes. It definitely sounds impressive. I've already got Illustrator but would be willing to give another program a shot if it can be replaced with something speedier and more convenient.

Inkscape's "Shapes" support is something I've wished Illustrator had for years, as well as the ability to manipulate gradients on the canvas instead of only within some slow-responding dialog/palette.

The graphics and visual layout of my site were done 100% in Illustrator, including their cutting, rasterizing and saving. I then run all PNGs through PNGCrush and PNGCrusher to make them smaller and strip out all metadata, including gamma information. These types of programs are far more well suited for web work than something like Photoshop, so any chance to improve on AI is welcome as well.

Actually i don't know all the ins and outs of inkscape, it justed worked nicely for the occasional work i needed a vector program for.
As far as i can tell, the export bitmap function only exports to png, and offers no control over bitdepth.

Photoshop CS3 has lots of new non-destructive editing capabilities and includes the ability to resize multiple times, always using the full data. You need to specifically enable this for each layer though and your PSD files will of course end up significantly larger.

Unfortunately that Inkscape export dialog doesn't hold a candle to Illustrator and Photoshop's "Save for Web."

Wen I finally have the chance to take a look at it, perhaps they'll have something a bit better. I might also be able to save out the artwork and load it into Illustrator for exporting. Though I'm looking to streamline the whole creative process not convolute it any more.