Is upsizing/zooming more effective when using "simple" ratios?

[xpatUSA]

Yes, this is what I meant by "needing a magnifying glass" to see the difference. Since a 50% zoom is "pure", I think it absolutely must be superior to similar zooms which are forced to do extra processing. But the difference should be small, given good software.

It seems that your understanding of re-sampling may be different to most, judging by: "similar zooms which are forced to do extra processing". As far as I know, no "extra processing" is done by any re-sampling algorithm for any reason. For that matter, 50% from the point of view of processing is no different than any other percentage, i.e. the processing does not change even though the result may be "better" or "worse".

As to the magnifying glass: at 400%, I still see no difference between 50% and 49%:



Have you tried my image in MS Photos yet?

[TomMartin]

Photos handles the outer part as expected but creates a variety of patterns in the interior section as the zoom level changes. It's acting as I described above - certain zoom ratios are more faithful than others. Since Photos doesn't tell you what the zoom level is I can't tell which are more accurate.

When I use FastStone the behavior is similar but the patterns are different. There are clearly "sweet spots" in the zooming. Some zooms are accurate while less intense zooms are not.

The more processing a program must do, the less predictable the results. So what could simplify a program's task? We know that images can be zoomed 50% or 200% blindingly fast so those are using minimal processing. As expected, the 50% zoom produces a very strong image when viewed at full size. The same cannot be said for other zoom levels although 49% does seem to do equally well. I'm not surprised that I can't predict which "complex" zooms will be accurate but at least all the "simple" zooms look good. That's not proof, but it certainly supports my argument.

All I'm looking for is a way to avoid the unpredictable variation in the center as you move through different zoom levels. I suspect it can be done by using simple ratios but I suppose I'll need to resize the image a few dozen times to generate enough data to prove my theory.

[TomMartin]

Here's something interesting - Photos shows a big difference between FastStone's 49% and 50% zooms. It shows the same distortions as FS does for the 49% zoom but displays a flawless 50% (much better than FastStone's rendering).

Edit: This is actually intermittent! Sometimes Photos' "full size" function shows it flawlessly, sometimes like FS. In the latter case I can largely remove the distortion by zooming up or down (just as with FastStone).

[xpatUSA]

Photos handles the outer part as expected but creates a variety of patterns in the interior section as the zoom level changes. It's acting as I described above - certain zoom ratios are more faithful than others. Since Photos doesn't tell you what the zoom level is I can't tell which are more accurate.

When I use FastStone the behavior is similar but the patterns are different. There are clearly "sweet spots" in the zooming. Some zooms are accurate while less intense zooms are not.

The more processing a program must do, the less predictable the results. So what could simplify a program's task? We know that images can be zoomed 50% or 200% blindingly fast so those are using minimal processing. As expected, the 50% zoom produces a very strong image when viewed at full size. The same cannot be said for other zoom levels although 49% does seem to do equally well. I'm not surprised that I can't predict which "complex" zooms will be accurate but at least all the "simple" zooms look good. That's not proof, but it certainly supports my argument.

All I'm looking for is a way to avoid the unpredictable variation in the center as you move through different zoom levels. I suspect it can be done by using simple ratios but I suppose I'll need to resize the image a few dozen times to generate enough data to prove my theory.

Hello Tom. I'm beginning to think that you may have never heard of moiré. These patterns that you see in the center are caused by it and can be predicted mathematically i.e., they are not unpredictable. The image I posted for you is unique in that the spatial frequency of the lines increases toward the center. Any zooming out causes aliasing somewhere in the image which shows itself as Moiré patterns - which look pretty because of the afore-mentioned varying of spatial frequency from the outside to the inside.

Here's some light reading on the subject:

https://photographylife.com/what-is-moire/

https://en.wikipedia.org/wiki/Moir%C3%A9_pattern

The effect can not be avoided. As you have observed, it is less noticeable when re-sizing by factors of two but there will still be aliasing - often showing up as "jaggies" on edges, straight or curved.

Not sure I can say any more about it, so good luck with MS!

[TomMartin]

Yes, I'm only vaguely familiar with that - thanks for pointing out that its effects do not increase proportionately with zoom.

So does anyone have a test image that would work better? I need something that you would expect to have uniform degradation the further the picture is zoomed from full size, without having to determine whether moire is responsible.

[xpatUSA]

Yes, I'm only vaguely familiar with [moiré] - thanks for pointing out that its effects do not increase proportionately with zoom.

Just for the record, that is not what I said. Open my image in FastStone and zoom out to an image size of 2% by pressing "-" repeatedly. The moiré will expand for each press until it's all over the image.

So does anyone have a test image that would work better? I need something that you would expect to have uniform degradation the further the picture is zoomed from full size, without having to determine whether moire is responsible.

Here's one:



I have posted it so that you can see the relationship between spatial frequency and aliasing. As you zoom out, there will be no aliasing artifacts appearing at all.

[TomMartin]

I know you didn't explicitly state "not ... proportionately" but I wanted to give you credit for what seemed an obvious conclusion. Was I mistaken?

Thanks for the new image but neither Photos nor FastStone can display it clearly on my machine. The good news is that it does behave as expected with no unexpected changes in clarity from one zoom to the next. I'll go through my images and try to find something that looks sharp at full size but deteriorates badly when moderately zoomed. Then I'll try zooming at more "basic" intervals to see if it makes a difference (I just wish I could do it with Photos, not FS).

The big problem is that I can't really trust the renderings - Photos (on a good day) says there's a substantial difference between 49% and 50%. FastStone says there's nothing that the unaided eye can see. It's not clear which program to trust.

To make sure we're on the same "moire" page, here are the differences between the 49% and 50% zooms.

[xpatUSA]

I know you didn't explicitly state "not ... proportionately" but I wanted to give you credit for what seemed an obvious conclusion. Was I mistaken?

Yes.

Thanks for the new image but neither Photos nor FastStone can display it clearly on my machine.

That's because the shot was out of focus.

The good news is that it does behave as expected with no unexpected changes in clarity from one zoom to the next.

As expected indeed.

[rpcrowe]

THIS STATEMENT REALLY TORQUES ME OFF (MY WIFE WOULD BE PROUD THAT I DIDN'T USE THE SLANG WORD FOR URINATION)...

"One thing I enjoy more about web images than published photos is that you can edit them however you like, producing a nicely customized collection that you truly did help to create. But I know you guys did all the hard work - thanks!"

I don't mind anyone grabbing my image from the web. I would rather they give me credit but, that doesn't matter. What would get me riled is that you would have the audacity to steal my image (or anyone's image) and then edit it! That is just morally not right - never mind if it is legal or not legal. Did you ever think of requesting permission or is Microsoft so big and powerful that you don't have to be polite...

[Polar01]

Richard I do not see how your post #29 fits into this thread about upsizing/zooming.

Cheers: Al

[Manfred M]

My understanding is that copyrights just prevent you from redistributing the images.

Not at all correct. Copyright means that the images belong EXCLUSIVELY to the creator and you can't do anything with them or to them (including saving them to your machine) or to them without the copyright holder's agreement. Editing them is definitely not allowed without explicit permission of the copyright holder either.

[Manfred M]

Richard I do not see how your post #29 fits into this thread about upsizing/zooming.

Cheers: Al

Have a look at entry #17.

[Polar01]

Sorry Richard and thanks Manfred missed entry #17, I must have looked a couple of times through or it is just too late and pass my bed time. Think I will go and warm up some milk.

Night, night: Allan

[Mike Buckley]

This is a simple Q&A with an attorney specializing in U.S. copyright law pertaining to photos.

The following is an excerpt that addresses probably the primary issue raised in this thread by Tom in his thinking that he has the legal right to alter images made by others (he doesn't assuming he is not making a bona fide derivative work):

"Q: What is a derivative work and who owns the copyright?

A derivative work is one that is based on one or more earlier works. Derivative works include editorial revisions, annotations or other types of modifications. The work must be different enough from the original to be regarded as a new work—in other words, it must contain some substantial, not merely trivial, originality. The threshold for originality in a derivative work is higher than that required for the original work.

The person who creates the derivative owns the copyright to revision, annotation, or other type of modification only. The original copyright is still owned by the original creator."

[Manfred M]

Mike - I think the important part of the article includes part of the material at the beginning of it:

"In simple terms, copyright for photographers means owning property. With ownership, you get certain exclusive rights to that property. For photographic copyrights, the ownership rights include:

(1) to reproduce the photograph;

(2) to prepare derivative works based upon the photograph;

(3) to distribute copies of the photograph to the public by sale or other transfer of ownership, or by rental, lease, or lending;

(4) to display the photograph publicly;"


In order to create a derivative work, the person needs permission to alter the work in the first place.

The issue, of course, is that the situation is complicated by geography. Copyright and other intellectual property laws vary from jurisdiction to jurisdiction, especially when it comes to the "fair use" provisions. In Canada, the interpretation of fair use for educational purposes is quite broad, based on a relatively recent Supreme Court decision. I'm fairly certain that downloading a file and editing it to practice my editing skills would be legal. I suspect I could also post it to show off what I had done, but I'm also fairly certain I could not pass off the work as being my own or sell it.

[TomMartin]

Sorry - I never realized some people felt that way about their pictures. It sounds like you only put "finished" pictures out there but there are a huge number of potentially great shots that only need a little editing to bring them out. Those are the ones that interest me the most - I like creating beauty. It's hard for me to see the harm if no one else ever sees them, but I think I understand where you're coming from.

[Mike Buckley]

Agreed, Manfred. When I wrote that Tom must be making bona fide derivative works for them to be legal, that means that they are truly bona fide, not made fraudulently or with deceit. In other words, bona fide derivatives can not be done without an explicit agreement with the original's copyright owner unless the original is in the public domain.

[Mike Buckley]

It's hard for me to see the harm if no one else ever sees them, but I think I understand where you're coming from.

When referencing the people who do what you do, you explained that "the most skilled have awe-inspiring collections." How do you know that if you haven't seen their collections? And if you've seen their collections, why are you saying that no one else ever sees them?

[xpatUSA]

Sorry - I never realized some people felt that way about their pictures. It sounds like you only put "finished" pictures out there but there are a huge number of potentially great shots that only need a little editing to bring them out. Those are the ones that interest me the most - I like creating beauty. It's hard for me to see the harm if no one else ever sees them, but I think I understand where you're coming from.

@TomMartin,

You could be alienating members, as you continue to provide seemingly oblique responses to well-intended comments. For example "where you're coming from" is a subtle shift, implying that someone was passing a personal opinion whereas in fact they were probably quoting civil law.

Not that I can be sure to whom "you're" is referring, of course.

We don't know what civil law applies specifically to you because you haven't given us your location or, for that matter, your real name which might or not be "Tom Martin".

So, your credibility could well be declining here and responses to your good self may well get less helpful.

For example, see my somewhat less-than-helpful post #28.

Just a heads up, not intended to be aggressive, and I am not a Moderator . . .

[tclune]

I have been retired for a few years now. But, when I was working for a living, my field was medical imaging software. One of the issues that we faced in the early days of my career was a problem of artifacts in images when they were resized. Of particular concern was nuclear images, like those from a gamma camera. These images are from noise sources, so they look like a B&W pointillist painting. When you resize such an image by a small amount -- say increase or decrease the size by 10% or so -- you would get what looked like a plaid pattern across the image when using a naïve algorithm (any variation on interpolation without prior smoothing would do.) The problem was that the interpolated data would blend neighboring points with greater or lesser weighting -- with a simple bilinear interpolation, for example,some interpolated pixels might be exactly 1/4 of each of the nearest neighbors and others might be just a few percent of three of the neighbors and almost entirely the fourth. So what? Well, noise is diminished at the square root of the number of samples, so the evenly weighted pixel would have half of the noise of the one that was essentially made up of only one pixel. The image would go into and out of "noise-smoothedness" in a regular way across the image, which created that irritating plaid artifact. The answer is clear enough -- low-pass the image to get the frequency content of the image below the Nyquist limit for the smaller image before resampling. And by the way, using extremely noisy images is a good way to test the integrity of your resampling algorithm. HTH.