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

[TomMartin]

Over at answers.microsoft.com people are wondering why certain zooms are more blurry than others, even though the clearer ones are often more heavily zoomed. For instance, a 70% zoom can be more blurred than a 90% zoom of the same picture.

My theory is that zooming/resampling is best done in fixed increments. So a 50% zoom will always look better than a 49% zoom.

But many hours of research has yielded nothing concrete. Before I suggest to Microsoft that they abandon their 10% zoom increments I want to know if I'm right that they're not optimal. Here are the ratios I'm about to suggest: 1/6 1/5 1/4 2/6 2/5 3/6 3/5 4/6 3/4 4/5 5/6.

[Manfred M]

Welcome to CiC Tom.

Frankly I have absolutely no idea what you are on about. I assume this has to do with a specific piece of MS software?

With the tools I use in digital photography, mostly Adobe Photoshop CC 2017, I've never run across anything like this. If there is an issue, it is likely related to the way the scaling algorithms have been designed as well as the underlying data that is being manipulated.

[Mike Buckley]

Just to confirm the potential confusion, I agree with Manfred that I have no idea what you are referring to.

[TomMartin]

Just to confirm the potential confusion, I agree with Manfred that I have no idea what you are referring to.

Thanks for both quick responses! The product is Microsoft's Photos and I imagine that will draw a few giggles. We're working to make it a respectable photo app, yet aimed at the masses.

The problem is that some of the user community is trying to understand the inconsistencies of zooming. Apparently this isn't a problem with most editors but I'd assumed that Microsoft was using fairly standard techniques (although naturally the most basic of them).

I didn't understand why there would be zooming "sweet spots" (even with old techniques) so I did some research and saw a few warnings about only upsizing in "even" intervals (25%, 50%, 75%...). The explanation given was that it was easier for software to change four pixels to five (25% zoom) than to change 25 pixels to 31 (24% zoom). No one disagreed with the statements so I accepted them because it explained what we were seeing and made a certain amount of sense.

I'm sure a 50% zoom must be better than a 49% zoom but to what degree? You'd probably need a magnifying glass.

So are techniques now so sophisticated that there's no need to worry about simplifying things for the software? If so, was this a problem in the past? I know MS won't change its zooming algorithms but it might zoom in "basic" intervals if we can make a good case for it.

[Mike Buckley]

The product is Microsoft's Photos...We're working to make it a respectable photo app, yet aimed at the masses.

It's just my opinion, but I have to ask why? There are already so many respectable photo apps aimed at the masses. It's not as if the industry needs another one trying to play catch-up.

[TomMartin]

Microsoft is using it to sell space on their OneDrive product. From the users' perspective, it used to have a good blend of simplicity and effectiveness. But now MS is shifting focus and working to make it something kids will like.

I now use FastStone but still do most of the support for Photos.

[Manfred M]

Tom - as you have suggested that MS is unlikely to change the algorithms, I suspect that you are in a no-win situation.

The issue is reassembling the data in a way that deals with the artifacts of the re-sizing process. There was a time that even Photoshop was not very good at this and the farthest I would go was a 30% increase in image size. After that softening from interpolation really got out of control. On One software's Perfect Resize as probably the best workaround some 10 years ago. My understanding is that they used fractals to overcome some of the softening issues.

So there are answers to these issues, but I would suggest the MS would have to undertake an exercise to redesign the resizing algorithms to overcome this softening issue. If they don't the issues you have identified will continue.

On the other hand, there are any number of excellent products out there that handle this issue well, so other than MS Photo being part of the package that ships with the OS, I see no reason why anyone who is serious about photography would stick with that particular product.

[TomMartin]

Thanks - I appreciate the insight. I don't want to make it sound like Photos zooms badly - its idiosyncrasies aren't often apparent. I was just hoping for an easy way to avoid the more complex parts of the program. So the people that warned not to upsize by random amounts were mistaken? In other words, it doesn't help to break the job down into smaller increments? It does seem like it would make the process more reliable but if no one has done such a study I imagine the benefits are too small to notice.

[Manfred M]

Thanks - I appreciate the insight. I don't want to make it sound like Photos zooms badly - its idiosyncrasies aren't often apparent. I was just hoping for an easy way to avoid the more complex parts of the program. So the people that warned not to upsize by random amounts were mistaken? In other words, it doesn't help to break the job down into smaller increments? It does seem like it would make the process more reliable but if no one has done such a study I imagine the benefits are too small to notice.

Given the audience for MS Photos, do you really think this matters? The audience is likely someone who is trying to do some fairly simple editing with what they already own via the OS. By and large, the images will have some fairly basic adjustments made to them and shared with family and friends, either directly or on screen or possibly printed on a basic colour printer or at the local photo printer found at Costco or Walmart. In other words, look at the target audience for this product and ask yourself, "is it good enough?".

If these people use anything other than the camera in their SmartPhone, they will also have a basic editor supplied by the camera manufacturer. Other options include some of the excellent on-line tools. If they get "serious" about it, they could download some of the excellent paid applications or free ones like Gimp.

I honestly think that the audience you are trying to address here will either be happy with the basic editing they get out of MS Photos and will never notice the minor issues you have written about or have a host of other options out there if they do.

[TomMartin]

You're certainly right that this is an issue MS is unlikely to address due to its target audience. For me it's been more of an intellectual exercise in trying to understand the problem. I thought if there was an easy way to make zooming more reliable, Microsoft might listen. And if not, at least I learned something that helps upsize web images.

[Manfred M]

You're certainly right that this is an issue MS is unlikely to address due to its target audience. For me it's been more of an intellectual exercise in trying to understand the problem. I thought if there was an easy way to make zooming more reliable, Microsoft might listen. And if not, at least I learned something that helps upsize web images.

If you look at how digital cameras and raw convertors work, you have something of a parallel situation. One can take a cluster of RGGB and turn that into a single data RGB point or one can take a cluster of data points around the RGGB grouping and generate a single RGB value from that. There are tradeoffs and artifacts that are outcomes of these technical decisions.

As MS primarily produces an OS and productivity software (MS Office), they have a totally different audience than companies like Adobe (imaging) or Corel (imaging and productivity). Some using PowerPoint will have a different audience than someone using Photoshop or Adobe Illustrator....

For me, upsizing is an issue I might have to deal with when upsizing an image to create a large print. but under normal cirucmstance, my 2.5K display requires that I downsize image from either a 36MP or 18MP camera to fit on the screen. My issues are more of a concern of getting the right colour space and the right amount of sharpening (downsizing inherently increase image sharpness) than issue associated with upsizing (softness due to interpolation) unless I work on really large prints.

[TomMartin]

A constant source of frustration for me is that none of the products are concerned with people with my specialty - collecting web images. We mainly need great upsizing but have little use for the bells and whistles (yes, I understand that we're Philistines). The best approach might be to collect images for a year or two and then get a demo copy of Perfect Resize. Instead I try to persuade companies to consider our needs (no luck so far).

So I guess what I'm looking for are published results of studies of best upsizing techniques (programmatic or user-oriented). I imagine most of those are proprietary but I was hoping that an old study or two may have found its way to the web.

[Mike Buckley]

How do you use your specialty of collecting web images?

[TomMartin]

Collecting via the web allows us to develop a very strong collection of photography in a very short while. Although the top 1% or so of images are beyond our reach, everything else is out there for the taking. Rather than concentrating on perfecting small numbers of photos we go for quantity and convenience.

That makes us sound uncultured and most of us are, but the most skilled have awe-inspiring collections (of regrettably small images).

[Round Tuit]

...Although the top 1% or so of images are beyond our reach, everything else is out there for the taking. ...

Are you implying that only 1% of images on the web are copyrighted or are you saying that you ignore copyrights.

André

[xpatUSA]

I now use FastStone but still do most of the support for Photos.

Now that got my attention, I use it myself. With the right settings, it is a useful diagnostic tool.

Amongst other things, you can enter integer zooms in the little box instead of selecting from the drop-down or clicking on the +/- buttons.

By entering your own fractional zooms as integer percentages, you can do your own research. You can turn off the smoothing for zooming and, for the smoothed zooming, you can select the re-sampling algorithm as well, IIRC.

Try this image for your tests; it can tell you much more than shots of shrubbery and rolling hills:


.

[TomMartin]

My understanding is that copyrights just prevent you from redistributing the images. The 1% figure was a rough guess as to the quality available on the web. I have dozens of high-quality photography books but I would say that 99% of the images have their equivalents on the web for free download. The only difference is size but even that is steadily becoming less of an issue.

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!

Finally, I'm not at all sure this bodes ill for pro photographers. Reviewing my collections makes me appreciate high quality photography more than ever; I still buy books not only for the increased size but to help support the art.

[xpatUSA]

My theory is that zooming/resampling is best done in fixed increments. So a 50% zoom will always look better than a 49% zoom.

My test says they look the same . . .





Resized in FastStone Viewer using Lanczos3 re-sampling.

[TomMartin]

Thanks for the FastStone suggestion, Ted. That was my next step if I received any encouragement that I might be on to something. The only problem is that it might well yield different results than Microsoft Photos. I know the latter is considered a joke by serious photo buffs but it's what I (voluntarily) support and certainly needs a lot more help. I agree that it's a questionable use of my time.

But if I see any reason to believe that upsizing can be optimized I'll use FastStone to see if I can draw any conclusions.

[TomMartin]

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.