Image resizing methods. Bicubic sharper disappoints.

[RichieRich]

If there's one thing which has always disappointed me about Photoshop CS3 it has to be its image resizing capabilities. In particular, I've never managed to get a decent result from the 'bicubic sharper' option, which is supposedly best for reductions. Today I decided to make some samples. These all pretty much reflect what I've experienced in the past.

Original image is a 4752 x 3168 RAW file taken with a Canon 50D. Images here resized to 640 x 427 pixels.

1) Canon DPP software with default settings (sharpness of 3)

The following three images were created by the following workflow:

- Open in Camera Raw (default settings)
- Open in Photoshop
- Resize (no pre or post sharpning)

2) Bicublic Sharper. In my opinion this consistently produces by far the worst results.

3) Bicubic. Not great, but with a little sharpening it's usable.

4) Bilinear. The best result which Photoshop achieved, but in my opinion still falling slightly short of the image which Canon's DPP software created. Will respond well to a bit of post-resize sharpening (not carried out on this image).

Naturally these are very basic workflows, however, I've never managed to create a workflow which involves resizing with Photoshop or Camera Raw which matches the results offered by Canon's sofware. This is disappointing as Camera Raw naturally offers certain advantages in terms of image manipulation and procesing which DPP simply doesn't have.

My usual workflow usually involves producing a resized image via Canon's sofware before processing the image in Photoshop and saving for web.

 

[RichieRich]

Anybody have any views on this? It'd be interested to learn what resizing options / workflows etc people are using.

 

[swanseamale47]

How does it compare with the DPP shot when theres no sharpening applied? and have you tried the PS versions with some sharpening? Also is DPP resizing from the raw rather than from an open image which I assume your doing in photoshop.

 

[RichieRich]

Yes, DPP is resizing the RAW file. Photoshop is resizing the file opened from Camera Raw.

DPP is still marginally better than bilinear in Photoshop when sharpen is set to zero in my view, but it's debatable. It also responds well to post-resize processing as it seems to retain the most detail.

What's surprised me is just how bad bicubic sharper is. It's seems to compare very poorly with all other algorithms for most pictures (with the exception of 'nearest neighbour'). I'd best describe the result as 'bubbly', for want of a better word.

 

[tiler65]

I don't know if it is my eyes but I think #2 is the best.........

 

[RichieRich]

I don't know if it is my eyes but I think #2 is the best.........

You're kidding me. Look at the trees to the left and in the background. They look like they're made of bubble wrap. Detail is only really retained in the first and last images.

Still, goes to show just how subjective this stuff is.

 

[Photodiva]

Hi. I've never resized this way. I usually go File-Automate-Fit Image, choose 800x600 or whatever, then apply a Filter-Unsharp Mask. Is there a reason you've chosen this method?

 

[DekHog]

I'd say it's pretty much impossible to tell viewing images at this size online anyway... they all look much the same to me. :shrug:

 

[RichieRich]

Hi. I've never resized this way. I usually go File-Automate-Fit Image, choose 800x600 or whatever, then apply a Filter-Unsharp Mask. Is there a reason you've chosen this method?

I guess I've used this method because it's called the 'resize' function. I've just checked your method and it appears in the history as 'resize image', the same as the method which I'm discussing here. My guess is that it's the same function but with the default compression algorithm applied - Bicubic (smooth gradients). This is used for image 3 in my post.

You may find that doing a resize with bilinear followed by a bit less sharpening may produce better results. Still, I'm yet to find a workflow which I'm really happy with.

 

[RichieRich]

I'd say it's pretty much impossible to tell viewing images at this size online anyway... they all look much the same to me. :shrug:

But you must be able to see that there's a huge difference between images 1 and 2 in terms of detail.

 

[Photodiva]

I'd say it's pretty much impossible to tell viewing images at this size online anyway... they all look much the same to me. :shrug:

I have to say that at the screen size here they appear very similar. My own experience is in submitting 1024x768 images for digital projection is where the problem appears greatest. On a monitor its hard to see exactly, but when they are blown up huge over-sharpening or lack of is really more apparent.

 

[tiler65]

You're kidding me. Look at the trees to the left and in the background. They look like they're made of bubble wrap. Detail is only really retained in the first and last images.

Still, goes to show just how subjective this stuff is.

I am not kidding, sharpening affects the edges that can be read by the software. Taking pictures of gorse/sea/sandy beaches will not help to determine sharpening.

look at the large rock in the foreground, the sharpening is better on that in shot 2 than in shot 1. Simple.

Take a shot of newspaper print and post that with the same methods.

 

[DekHog]

But you must be able to see that there's a huge difference between images 1 and 2 in terms of detail.

The second certainly looks better, but also looks like more of a colour/contrast/saturation change rather than anything else. Maybe slightly sharper, but like I said, difficult to tell at this size.

 

[RichieRich]

This is really interestng. To me the second shot looks worse in every respect except colour. The detail in the telegraph poles in the background has been retained with the first but is rather smudgy with the second. If you inspect the images closely as I have you can see that much of the detail has been wiped out of the second one.

Still, I can't argue with what you find more pleasing to the eye.