Keeping Image Quality after Shrinking

[Mote]

Hello, i'm trying to shrink a triangular image using the "image size" tab. However, after shrinking the triangle edges become jagged and the text within it becomes blurred and uneven. Any suggestions at keeping the quality of the image after shrinking? I've kept the "constrain proportions" checked and i've changed the image to RGB mode. Thanks for your help!

[Mote]

Hello, i'm trying to shrink a triangular image using the "image size" tab. However, after shrinking the triangle edges become jagged and the text within it becomes blurred and uneven. Any suggestions at keeping the quality of the image after shrinking? I've kept the "constrain proportions" checked and i've changed the image to RGB mode. Thanks for your help!

[Beth]

Hi Mote

The fuzziness is part and parcel of any resampling (resoloution change). And probably, a triangle will be the hardest thing to resize due to the sharp angles involved.

There are a few things you can do to help minimize the problem though. First off, make sure your interpolation method is set to bicubic (in general prefs). Avoid resizing over and over again. This will just compound the problem...if you do not get the adjusted size the first time around, revert or undo and try it again. (And if you are just resizing part of a layer...it might even be handy to copy that layer for safe keeping...just incase you need to resize later on.)

After resizing the image, you can get some of the sharpness back by running the unsharp Mask filter a time or two. Becareful you dont do this too much as it can make undesirable noise more prevalent or make image 'pop' out too much.

I believe that resizing is best done 1/2, 1/4 or 1/8 (as well as doubling quadrupling etc). This is a matters of some debate however. But I believe that Photoshop handles these conversions better than say a .0426 increase in size. But a triangle, is going show you problems no matter what. That is why I always do a prescan for position only when doing my rough layout jobs. When the design in completed, I scan in the photos at exactly the size I need for the final print job.

Beth

[BR]

Great post/reply Beth...... I haven't seen any debate on the subject (wish I did), nor even mention of it before, so I'd like to stress it.

I totally agree with you. If I need to resize at all, I always get the best results by using even fractions (or multiples), i.e. 75%, 50%, 25%, 12.5%, 6.25%. Could this be a bicubic thing?

BR

[Jen Worden]

I concur ... I've had best results with "even" resizing as well.

I also find that I get a better result when I don't use the "Image Size" to reduce image. I change it to a layer and then use "Edit > Transform" - generally using "Scale" but on occasion have tried "Numeric" with varying luck.

As with Beth, try and get the size correct the first time - Ctrl-Z and re-do if necessary. After which, I prefer "Filter > Sharpen" and then reduce Sharpen by about 50%.

Hope that helps! (Hi Beth! [img]/infopop/emoticons/icon_smile.gif[/img] )


cfn ... Jen
Jen Worden
Web Developer
www.meadoworks.com

[Beth]

Well thanks BR. We aim to please around here [img]/infopop/emoticons/icon_wink.gif[/img]

Yeah, the resizing thing seems to make sense doesnt it. But it might just be that my math education is drilled into my brain too much. 10 divided by 2 makes 5, but 10 divided by 3 equals 3.33. You look at those equations, and you see it is easier to divide by 2 than 3. But does photoshop think this way. Maybe not...I was just doing a couple of tests using a 300 by 300 image with an 100 by 100 black square on a white background. Okay, yes this is overly simplified...but it shows something interesting.On doubling or halfing the image...you end up with fuzzy edges around the black square. A one pixel margin of pixels outside the square are light grey and another margin inside the black are a dark grey. Not what we mathmaticians would expect. I mean arent mathmaticians black and white kinda people. BUT, when you try resizing by 137% (or any other weird number)...you get the same results...one pixel of liter grey and 1 of darker. Although, in either case, running the unsharp mask filter fixes the problem.

Hey Jen. Long time no hear.
I imagine the sharpen filter works better for low res web work. But it really is a fairly simple filter. There are no settings for the user to imput. It essentially just increases the contrast between edge pixels. The USM (unsharp mask) does SO much more. For one thing, you have control over various settings by using radius, amount and threshold. By playing around with these (and it all depends on the image), you a great deal of control as too what pixels will be affected, how many and how intense the effect will be. Photoshop does this by combining a slightly blurred (unsharp) image with the original and does some fancy math based on the setting the user requested. The end result is that a few pixels (based on radius setting) on the lighter side of a border will be even lighter than they were in the original. Those on the darker side, will be darker than the original. When viewed from a normal viewing distance for print work...this defines the edges better than a simple sharpen will. But for web work, this halo of lighter and darker pixels can be undesirable. Especially since you are dealing with so few pixels to begin with.

River Derchi,
Beth

[Jen Worden]

Guess I should have mentioned that I was thinking about images for the web. Just shows where my mind is at. Thanks for the great explanation of USM, Beth!

cfn ... Jen
Jen Worden
Web Developer
www.meadoworks.com

[Peter Clifton]

Hello people

In one of my computer arts magazine there was a small tutorial on resizing images, particularly those produced by Bryce.

Reducing a 1024x768 image to 320x200 in one step, was too much for an photo editing program's resampling engine to handle and a lot of detail was lost.

Their advice was to never reduce such images by more than 50% in any one step, and prior to the second resizing stage use the unsharp filter to attempt to restore some of the detail.

In the examples given there was a noticeable improvement in the two stage reduced image as opposed to the one stage reduced image.

Peter


[This message was edited by Peter Clifton on September 13, 2000 at 08:45 AM.]

[Beth]

Well, I did a few tests again. I did them with photos for print work....so there was considerable pixels to play with. And you are right, the results were better if resizing is done in increments no larger than 50%.

I did not test web res stuff though. I think any undesirable anti-alliasing would show up there.

My moto...create your tiffs at the size you need

clients here gotta go