DPChallenge: A Digital Photography Contest You are not logged in. (log in or register) 
 

DPChallenge Forums >> Photography Discussion >> Resizing ... grumbles from a mad man.
Pages:  
Showing posts 1 - 22 of 22, (reverse)
AuthorThread
 11/05/2010 09:22:08 AM · #1
Okay so.. I'm sure I'm not the first to notice, grumble, whatever about this.

But my current image on the free study that is currently still in voting is of course coming out lower than I would have thought.. Now to be honest after going through the free study I don't think I have a 6 here, but I was hoping for a more mid level 5 but I'm stuck at the 5.1.

I have a couple of theories here.

1) I just suck (completely possible).
2) Reducing the image down to 800xwhatever is killing a lot of detail.

As a test I used my wife who loves everything I do of course, (but that's not the point).

What I did is I showed her the same image, the one I turned in at it's 800xwhatever and a image that was double the size, so still reduced but reduced to 1600xwhatever.

Well it was pretty clear here that the 800xwhatever was losing a lot of detail that I think makes the photo what I think it should be getting. My wife even noticed that she liked the bigger one.

Again.. I could just suck at the photography stuff.

So the question is, do a lot of you spend time looking at your photos at the reduced resolution and end up changing your mind about turning it in?

Another question is, what are some things you do when you reduce to keep that extra bit of detail etc?

Message edited by author 2010-11-05 11:37:03.
 11/05/2010 09:25:04 AM · #2
I know it's probably not the smartest thing to do, but my first editing step for challenges is the reduction.
 11/05/2010 09:28:30 AM · #3
I have the following saved as an action in CS5 that I run before my final save:

Unsharp mask 18,40,0
Unsharp mask 150,.3,0 (note that is .3)
edit>fade unsharp mask 100% darken
Unsharp mask 150,.3,0
edit>fade unsharp mask 50% lighten

It brings back most of the detail and sharpens the reduced photograph.
 11/05/2010 09:33:05 AM · #4
maybe someone else can confirm this but when you resize to 800px the internet browser is displaying the picture at 1:1. so it doesn't matter if there are extra pixels in your original you aren't going to see them if they are he same relative size.

Probably not the best explanation.

11/05/2010 09:39:58 AM · #5
First of all, 800 pixels is better than 640 pixels. That previous small constraint was a severe handicap and it did favor more simple subjects with less detail. When 800 pixel images were authorized for all challenges, the quality constraint became the 300kb file size. I sure can't tell the difference between 85% quality JPG compression and 95% JPG compression. But, I can tell the difference between 66% and 95%. More complex, more detailed, more colorful, more sharpened images sacrifice loss of display quality when constrained to 300kb due to additional JPG compression. By definition, JPG compression discards like-neighbor pixels. Carried to an extreme, artifacts and posterization can occur.

There is the issue of small computer monitors and laptop monitors. Those legacy displays probably mean a pixel dimension allowance greater than 800 pixels is not in the cards. This is especially true for any greater allowance on the vertical dimension. It has been suggested by some that a constraint of 800 pixels on the vertical dimension be coupled with an allowance of 1200 pixels on the horizontal dimension. That would be helpful to more complex subjects. But, if the 300kb file size constraint stays in place, then a larger dimension allowance would not be an advantage.

I suppose there is the issue of bandwidth usage. Because of the membership size of DPC and the page views, the use of a lot more bandwidth would come at a cost. I'd be willing to pay a premium membership fee in order to be granted larger dimension images AND a higher file size allowance. But, then this wouldn't be an even playing field, I guess.
 11/05/2010 09:40:05 AM · #6
There are different resizing algorithms. Some of them do a better job than others. If you are using the wrong one, it can totally kill an image when you scale it down. I usually use bicubic, as it tends to produce a result with the most detail. Final sharpening is done with unsharp mask at a radius of .35 and amount of usually 150-200. Using a very small radius and large amount gives me good results without sharpening halos.

I've finished voting the freestudy. If you like, Ill be happy to give you further feedback if you PM me which entry is yours. I won't change any votes, regardless of what they are.
 11/05/2010 09:44:39 AM · #7
Originally posted by Aarthek:
My wife even noticed that she liked the bigger one.

That's to be expected.
 11/05/2010 10:01:20 AM · #8
Originally posted by JH:
Originally posted by Aarthek:
My wife even noticed that she liked the bigger one.

That's to be expected.


ROFL!

Resizing is an art. The goal is to keep as much detail as possible without visible artifacts. Try this:
1.) Edit the image at full size
2.) Sharpen the image as required, at full size
3.) Reduce the size to *twice* the final size; if you are submitting an 800px wide by 600px tall image, resize to 1600px wide by 1200px tall. Use "bicubic" resampling.
4.) Do your final resize, using "bicubic sharper."

You can also experiment with applying some moderate sharpening between steps 3 and 4, but use very small radius and moderate amount. Alternatively, some *very* gentle sharpening of the final image is sometimes beneficial. A lot depends on the image content and the sharpness of the original. There is no one canned process that will always yield good results.
 11/05/2010 10:14:49 AM · #9
Originally posted by kirbic:

Resizing is an art. The goal is to keep as much detail as possible without visible artifacts. Try this:
1.) Edit the image at full size
2.) Sharpen the image as required, at full size
3.) Reduce the size to *twice* the final size; if you are submitting an 800px wide by 600px tall image, resize to 1600px wide by 1200px tall. Use "bicubic" resampling.
4.) Do your final resize, using "bicubic sharper."

You can also experiment with applying some moderate sharpening between steps 3 and 4, but use very small radius and moderate amount. Alternatively, some *very* gentle sharpening of the final image is sometimes beneficial. A lot depends on the image content and the sharpness of the original. There is no one canned process that will always yield good results.


I'll try your method kirbic. Seems interesting and I'm confident it works well. Sharpening is otherwise the thing I save to the very last; just prior to saving the file. It works for me and I don't think I lose too much detail this way.

Lately though... I export to 'DPC format' directly from Lightroom and apply sharpening in the export process if necessary. For me this gives the best results. I have LR set in a way that it saves the longest edge to 800 pixels and that the final result is never to exceed 290kb. The file is then DPC-ready :)
 11/05/2010 10:20:46 AM · #10
Originally posted by TrollMan:


Lately though... I export to 'DPC format' directly from Lightroom and apply sharpening in the export process if necessary. For me this gives the best results. I have LR set in a way that it saves the longest edge to 800 pixels and that the final result is never to exceed 290kb. The file is then DPC-ready :)


I use the same method and it seems to work just fine.
 11/05/2010 10:35:06 AM · #11
You can try Adamus sharpening - reducing the oversharpened image can miraculously preserve a lot of details in final size image.

Here's a simplified version (I actually don't know how it was used originally):

- resize to 1600 px longest dimension (2-times the desired output size)
- apply Sharpen filter twice (Filter-Sharpen-Sharpen) - the image will look quite ugly at this stage
- resize to 800 px - miracle happens at this stage, a lot of fine details will be preserved.

If I don't use the above technique, I just resize image to 800 px and apply Unsharp mask with 0,2px/400%/0
11/05/2010 11:05:10 AM · #12
Originally posted by JH:
Originally posted by Aarthek:
My wife even noticed that she liked the bigger one.

That's to be expected.


+1 well played.
 11/05/2010 11:16:46 AM · #13
Losing. LOsing. One "o". Thank you kindly!
 11/05/2010 11:37:44 AM · #14
Originally posted by Melethia:
Losing. LOsing. One "o". Thank you kindly!


good lord.. really? one typo and ..

never mind.. I'll just end my original posting..
 11/05/2010 11:39:05 AM · #15
thanks all.. I'm going to try some of these great ideas.. thanks for the comments.
 11/05/2010 11:48:35 AM · #16
Originally posted by Aarthek:
Originally posted by Melethia:
Losing. LOsing. One "o". Thank you kindly!


good lord.. really? one typo and ..

never mind.. I'll just end my original posting..
nothing personal. Just a pet peeve. I'm on a campaign to eliminate the overuse of "loosing". Seems the thread provided very good tips. And yes it was worse when we had 150k size limits!
 11/05/2010 12:21:46 PM · #17
Originally posted by fridjo:
You can try Adamus sharpening - reducing the oversharpened image can miraculously preserve a lot of details in final size image.

Here's a simplified version (I actually don't know how it was used originally):

- resize to 1600 px longest dimension (2-times the desired output size)
- apply Sharpen filter twice (Filter-Sharpen-Sharpen) - the image will look quite ugly at this stage
- resize to 800 px - miracle happens at this stage, a lot of fine details will be preserved.


Adamus sharpening is generally excellent for DPC; I use it for all my submissions. But you're missing one, very useful, section:

1. resize to 1600
2. dupe BG, sharpen dupe
3. repeat step 2 twice

you now have 4 layers, no sharpening plus 1x sharp, 2x sharp, and 3x sharp

4. resize to 800

now start turning layers off from the top down, to find the amount of sharpening that works best for you. You can reduce opacity on the topmost saved layer if you want, say, 1.5x sharpening. When you have it nailed down, do your save-as.

Now, technically this is not legal for basic editing, but you can work around that by doing the above, making a note of how much sharpening you end up with, then reverting to step 1 in history and applying the correct amount of sharpening, as predetermined, from that point.

The point here is that different images require different amounts of Adamus to shine their best, and this system lets you compare very easily with the layer on/off switching.

R.
11/05/2010 01:57:59 PM · #18
One way I try to retain detail is by resizing by an exactly even (integral) amount. When resizing, the algroithm samples a certain number of pixels, then averages their value and assigns that to a new pixel. So, for example, my camera shoots at 2816 x 2112 pixels. If I reduce to 25% of the original, it will be 704 pixels tall -- a little smaller than the maximum allowed, but big enough for me -- and each pixel should have been derived from a 4x4 patch of pixels, with no need to calculate values for "fractions of a pixel." If I'm using a square crop, a reduction of 33.33% also gives me a 704 pixel image, with each pixel derived from a 3x3 batch of pixels.

I have no "proof" that this makes a difference, but it seems logical that the algorithm will work faster (if not more accurately) using reductions which use an even number of pixels (20%, 25%, 33.33%, 50%).

Another important factor is that your monitor can destroy detail if you view at the "wrong" magnification, such as using a "fit to screen"-type command. If your picture looks bad, try viewing at exactly 25%, 50% or 100% on your monitor.

Message edited by author 2010-11-05 13:59:10.
 11/05/2010 02:12:25 PM · #19
Originally posted by GeneralE:
Another important factor is that your monitor can destroy detail if you view at the "wrong" magnification, such as using a "fit to screen"-type command. If your picture looks bad, try viewing at exactly 25%, 50% or 100% on your monitor.


Absolutely. A lot of people are fooled when editing because they are using an interpolated size rather than a native size. ALWAYS study your edits critically at 100%, 50%, numbers like that.

R.
 11/05/2010 04:48:12 PM · #20
Originally posted by GeneralE:
...I have no "proof" that this makes a difference, but it seems logical that the algorithm will work faster (if not more accurately) using reductions which use an even number of pixels (20%, 25%, 33.33%, 50%)


The difference is not huge, but it is there. One key thing is that you only really need to worry about keeping the integer ratio on the last resize. Best practice is usually to resize no more than a 2:1 ratio in one step, using the first resize to get an integer multiple of the final size.
If your starting images are really big, you may want to resize in more than two steps, for instance:

4368x2912 ==> 3200x2133 ==> 1600x1067 ==> 800x533

Notice I used integer resizing for the last two steps.
 11/05/2010 05:30:56 PM · #21
Originally posted by JH:
Originally posted by Aarthek:
My wife even noticed that she liked the bigger one.

That's to be expected.


snap! lol

yeah, I used to get frustrated with the same thing.. now I do all my editing and what not, then save it, then save it as a new name, resize it in photoshop, and if it needs it run an unsharp mask, and save with highest settings you can. Also the poster above that had the unsharp settings that works really well too but it does change the contrast and look of it a bit in my opinion.. but usually for the better.. gives it a bit more pop

hope that helps
 11/06/2010 04:26:15 AM · #22
Originally posted by Bear_Music:
Originally posted by fridjo:
You can try Adamus sharpening - reducing the oversharpened image can miraculously preserve a lot of details in final size image.

Here's a simplified version (I actually don't know how it was used originally):

- resize to 1600 px longest dimension (2-times the desired output size)
- apply Sharpen filter twice (Filter-Sharpen-Sharpen) - the image will look quite ugly at this stage
- resize to 800 px - miracle happens at this stage, a lot of fine details will be preserved.


Adamus sharpening is generally excellent for DPC; I use it for all my submissions. But you're missing one, very useful, section:

1. resize to 1600
2. dupe BG, sharpen dupe
3. repeat step 2 twice

you now have 4 layers, no sharpening plus 1x sharp, 2x sharp, and 3x sharp

4. resize to 800

now start turning layers off from the top down, to find the amount of sharpening that works best for you. You can reduce opacity on the topmost saved layer if you want, say, 1.5x sharpening. When you have it nailed down, do your save-as.

Now, technically this is not legal for basic editing, but you can work around that by doing the above, making a note of how much sharpening you end up with, then reverting to step 1 in history and applying the correct amount of sharpening, as predetermined, from that point.

The point here is that different images require different amounts of Adamus to shine their best, and this system lets you compare very easily with the layer on/off switching.

R.


Thanks, this is a very elegant way to test different amounts of sharpening! I had my version saved as an action already from my pre-member times, when most of the challenges I submitted were Basic editing, so I just forgot about using layers.

Layers can also serve to have more freedom of adjusting Fade... after using unsharp mask - duplicate BG, apply unsharp mask and adjust Opacity for the top layer.
Pages:  
Current Server Time: 04/25/2024 01:33:23 PM

Please log in or register to post to the forums.


Home - Challenges - Community - League - Photos - Cameras - Lenses - Learn - Prints! - Help - Terms of Use - Privacy - Top ^
DPChallenge, and website content and design, Copyright © 2001-2024 Challenging Technologies, LLC.
All digital photo copyrights belong to the photographers and may not be used without permission.
Current Server Time: 04/25/2024 01:33:23 PM EDT.