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

DPChallenge Forums >> Hardware and Software >> Creating the sharpest image ..
Pages:  
Showing posts 1 - 9 of 9, (reverse)
AuthorThread
 03/14/2025 11:25:10 PM · #1
i dont ever crop the original canvas size on my photos ..
i duplicate the background layer and scale it up to fit the original canvas size ..
i often change the dimension of the original canvas size though .. by increasing the pxls width or length ..
hope that makes sense ..

anyway .. i was wondering if i'm sacrificing some sharpness and quality when i do it that way ..

i know there are some very knowledgeable ppl here that might know that answer ..

thankyou in advance .. :)
 03/15/2025 06:47:35 AM · #2
Not an expert here, but just making some considerations.

If you are adding pixels then this will always require an algorithm to guess/calculate/determine the pixels that need to be added and this will be based on the surrounding pixels.

When I started with Photoshop 20 years ago you had to take care of it yourself and the advice was to do the increase in small steps, otherwise the guess might not be as good and it would become noticable. Even at that time, 100% increase was possible. I can see over the years the algorithms have become much better and I don't see this to be a problem any longer. Also the results with new generative AI that can generate large new portions of an image without having surrounding pixels is really impressive, so generating new pixels in between should be "easy". I guess not only the immediate neighbouring pixels are taking into account but also larger areas.

The generated pixels might not be the same as what would have been there if you had higher resolution to begin with, so you might consider that a reduction in quality, but only if quality is defined as "representing the real world as accurately as possible". The recording itself and then the post-processing that is usually done already makes the image deviate from real life, so that would be a very narrow definition.

Also in sharpness, which can be seen as sharp edges, sharp intensity/color changes from one pixel to the other, you might loose something because the generated pixel will usually be somewhere in between. But sharpening methods have also improved.

It also depnds on your starting image. Upscaling compressed JPG image from my low spec phone definitely does not work. Upscaling RAW images from my camera does. Don't use compression with quality loss (JPG) if possible, and, if needed, only in the final step.

Basically, I would not worry about it. If it looks good, it is good. Maybe only if you want to make very large prints and you need to upscale a lot, make a small pre-print of a small area before doing the total job.

03/15/2025 08:14:30 PM · #3
Originally posted by willem:
Not an expert here, but just making some considerations.

If you are adding pixels then this will always require an algorithm to guess/calculate/determine the pixels that need to be added and this will be based on the surrounding pixels.

When I started with Photoshop 20 years ago you had to take care of it yourself and the advice was to do the increase in small steps, otherwise the guess might not be as good and it would become noticable. Even at that time, 100% increase was possible. I can see over the years the algorithms have become much better and I don't see this to be a problem any longer. Also the results with new generative AI that can generate large new portions of an image without having surrounding pixels is really impressive, so generating new pixels in between should be "easy". I guess not only the immediate neighbouring pixels are taking into account but also larger areas.

The generated pixels might not be the same as what would have been there if you had higher resolution to begin with, so you might consider that a reduction in quality, but only if quality is defined as "representing the real world as accurately as possible". The recording itself and then the post-processing that is usually done already makes the image deviate from real life, so that would be a very narrow definition.

Also in sharpness, which can be seen as sharp edges, sharp intensity/color changes from one pixel to the other, you might loose something because the generated pixel will usually be somewhere in between. But sharpening methods have also improved.

It also depnds on your starting image. Upscaling compressed JPG image from my low spec phone definitely does not work. Upscaling RAW images from my camera does. Don't use compression with quality loss (JPG) if possible, and, if needed, only in the final step.

Basically, I would not worry about it. If it looks good, it is good. Maybe only if you want to make very large prints and you need to upscale a lot, make a small pre-print of a small area before doing the total job.


thanks so much Willem for taking the time to give me all that information ..
unfortunately a lot of it went over my head ..
the generating pixels in particular .. how am i creating more pixels ..
do you mean when i scale up the layer i'm making the pixels bigger ie .. generating them .. so its like i have less covering the canvas .. in which case the quality isnt as good .. almost like the difference in quality between a camera with 12 megapixels and one that had say 24 megapixels ..??..

anyway .. i think the gist of your message is that no matter how we do it .. by cropping the canvas .. or by keeping the canvas size and scaling up the layer .. then its not going to be much different .. ?? ..

here's another one though ..
when i resize for dpc .. say i have an image that's 5000pxls width and i want to make it 1200pxls ..
i take it down in steps .. so i bring it down to 4000, then to 3000, then to 2000, and then to 1200 ..
i read it somewhere that it retains the quality or something ..
i'd be interested in what other people do .. and whether anyone thinks there's any difference bring the image down in steps or all at once ...

anyway .. thanks again .. and hope you are having a great weekend .. :)

03/16/2025 09:15:01 AM · #4
OK, maybe I made a wrong assumption. I thought you were changing image size in terms of pixels in order to be able to make a bigger print for example.

For example in Photoshop, if you are changing image size and have the "resample" box checked, then Photoshop will add pixels in between the existing pixels. Overall you will get more pixels so you can then print the same size at a higher pixel-per-inch count, or you can print a larger size at the same pixel per inch count. This was the basis for my first explanation.

A nice experiment is to start with an image of 2 by 2 pixels, so 4 pixels in total. If you make the top left and the bottom right pixels black (just select the pixels and fill them with black) and the other two pixels white then you can see what is happening.

Try increasing the pixels size of that image to 4 by 4 pixels and you will see that Photoshop generates new pixels. The way it does that depends on the resampling method chosen.
If you chose for example bicubic then it will new pixels with various shades of grey. If you choose nearest neighbour then it will create new pixels with the same color as the neighbouring pixel, so either black or white. In this example the nearest neighbour image will look the same as the original but the bicubic one will look very different with all different shades of grey. Your choice of resampling method depends on your starting image and the result that you want to achieve.

If you are only changing canvas size of the 2x2 image, let's say to 200 percent of original height and 200 percent of original width, then Photoshop will keep the original in the middle and just add additional pixels on the outside of the original and those new pixels will have the background color.

When I re-read your initial question it looks to me you are keeping canvas size (in terms of pixels and dimensions), but you are adding a new layer (can be copy of original) and then resizing that layer to bring your subject closer, to zoom in. Right? And then you might change the dimensions (like the ratio between height and width), so this last step is the same as cropping the image (unless you also use the transform tool, or scale tool, in which case also the proportions of the subject changes in the same way as the total image)

I think if you are resizing the new layer, zooming in (using the photoshop transform tool and pull the edges outwards), the software will do the same as with image resizing. If you again take the 2x2 image, or the 4x4 image that you have created with neirest neigbour resampling and you duplicate that layer and then zoom in (transform tool) you will see the same effect as resampling with bicubic, i.e. with shades of grey generated.

It probably also depends on your software. I never thought of using such a 2x2 image before but it works quite well to reveal what is happening.

If you crop an image, you just take a smaller portion of the total, so you are throwing away pixels that are outside the crop. You are not loosing any quality in the part that is remaining. As long as the final display and the final print does not enlarge the remaining part beyond its limits (for prints the recommendation is 300 pixels per inch) then all is fine.

Your question about resizing for dpc, I always do it in one step and then apply a unsharp mask after that. I would not expect a stepwise reduction to be better, does not make sense to me, but maybe I am wrong on that.

So, question remains why are you scaling up a duplicated layer instead of cropping an image? What are you trying to achieve?

Message edited by author 2025-03-16 09:18:43.
 03/16/2025 03:11:34 PM · #5
thanks again Willem ..

sorry i should have been clearer when i first asked my question ..

firstly i'm using photoshop ..

and to clear up what i mean by scaling up rather than cropping ..

i duplicate the background layer ..
i then select the free transform tool ..
i drag the edges out using the scale option ..
i drag it out to larger than the original canvas size ..
i can also rotate as well when i'm still in the free transform tool ..
i then apply ..

i then have the ability to move that transformed layer to where i want it to be within the canvas ..

i sometimes then adjust the canvas size .. by either increasing the width or length .. to make the image more pleasing depending on the original image ..
when i have adjusted the canvas size i can move that transformed layer around to see how it best fits the changed dimension ..
if i havent changed the canvas dimension .. having that transformed layer available makes it easy to move it around and see how it looks ..

if i had cropped the image i would have lost parts of the image that i'd have no way of getting back .. and i would have no way to play around with different looks ..

i'm hoping that makes sense ..

anyway .. thanks again .. i really appreciate the time you have taken .. :)
 03/16/2025 09:09:27 PM · #6
Scaling in any way will create new pixels as previously described.

It may be better to use the "Image Size" option as you can then set the resampling option -- usually "Bicubic Smoother" if enlarging and "Bicubic Sharper" if making smaller (as for resizing for DPC).

You might want to review this (old) tutorial on resampling.
 03/17/2025 04:52:26 AM · #7
I understand you want to look at the framing and proportions of the final image before throwing away any pixels outside the frame.

But is that not exactly what the crop tool allows you to do?

If you select the crop tool in photoshop and draw a suggested frame inside your original picture, then it will show you a light frame with darker borders.
The light part is the remaining image if you apply that crop, the dark part is the part you would throw away.

Before your apply this crop, you can move it around, rescale it, change the dimensions, rotate it, etc. just by dragging the edge handles or, for rotation, by moving your cursor outside the frame.
Only when you hit the enter key the crop is final. If you decide you don't like it after all you can always go back a step in the history and start over.

No loss of quality, just loss of the pixels outside the frame, but that was precisely your aim.

The remaining image you can then resize for dpc. Usually you will have much more than 1200 pixels left on the long edge.

Addition: you can even change the shield color and/or opacity in the top menu bar for the crop tool if you don't like the standard dark borders or want to switch them off completely.

Message edited by author 2025-03-17 04:56:15.
 03/17/2025 05:36:52 PM · #8
Originally posted by willem:
I understand you want to look at the framing and proportions of the final image before throwing away any pixels outside the frame.

But is that not exactly what the crop tool allows you to do?

If you select the crop tool in photoshop and draw a suggested frame inside your original picture, then it will show you a light frame with darker borders.
The light part is the remaining image if you apply that crop, the dark part is the part you would throw away.

Before your apply this crop, you can move it around, rescale it, change the dimensions, rotate it, etc. just by dragging the edge handles or, for rotation, by moving your cursor outside the frame.
Only when you hit the enter key the crop is final. If you decide you don't like it after all you can always go back a step in the history and start over.

No loss of quality, just loss of the pixels outside the frame, but that was precisely your aim.

The remaining image you can then resize for dpc. Usually you will have much more than 1200 pixels left on the long edge.

Addition: you can even change the shield color and/or opacity in the top menu bar for the crop tool if you don't like the standard dark borders or want to switch them off completely.


thanks again Willem ..

yes i know about the crop tool and how i can use it ...
ie. moving around etc etc before applying it ..
and i've played around with both cropping .. and my method of scaling up ..
but for me scaling up works so much better ..

i can go back to that transformed layer and adjust after i've applied the transform tool .. and adjust it again and play around with different looks because i havent 'deleted' any of the pic .. also .. i might have done a lot of other things after i've applied that transform tool .. like creating other layers etc etc ..

and excuse the fact that i might no be using the correct terminology ..
ultimately i suppose its different strokes for different folks .. !! .. ;)

anyway .. as always i appreciate the time you have taken to share with me all that information .. :)
 03/17/2025 05:38:36 PM · #9
Originally posted by GeneralE:
Scaling in any way will create new pixels as previously described.

It may be better to use the "Image Size" option as you can then set the resampling option -- usually "Bicubic Smoother" if enlarging and "Bicubic Sharper" if making smaller (as for resizing for DPC).

You might want to review this (old) tutorial on resampling.


thanks so much Paul ..
i will check that out ..
much appreciated .. :)
Pages:  
Current Server Time: 06/22/2025 01:36:35 AM

Please log in or register to post to the forums.


Home - Challenges - Community - League - Photos - Cameras - Lenses - Learn - Help - Terms of Use - Privacy - Top ^
DPChallenge, and website content and design, Copyright © 2001-2025 Challenging Technologies, LLC.
All digital photo copyrights belong to the photographers and may not be used without permission.
Current Server Time: 06/22/2025 01:36:35 AM EDT.