...now if you were to ask *me* (even though I'm not a pro) I would at least have a rational argument for why I am charging a client $10k for my work.

""Paul and klausesserare are absolutely correct in stating that upsampling an image will deteriorate the quality of an image, as the interpolation of pixels is at most clever, but is definitely non existent and can be seen, "once spotted never forgotten".""

And *that* is not a rational argument.

I can "see" the difference in "quality" between using a $900 50mm prime on a $2500 24MP dslr, and using a $50 5MP p&s.

Don't tell me that I can justify charging a client 15x as much because I use one camera vs the other.

I can throw that $50 5MP p&s into a motorized head and generate as many pixels as you want.

Can I then charge the client based on the MP of the final pano?

I can go onto the Internet and download fully-functional versions of C#, Silverlight, MySQL, Linux, Apache, JSEE6, ColdFusion, Flash and PS and build web-pages all day long. Can I then justify charging a client the full cost of functional licenses of all that software? No!

Could I even say that I *need* all that stuff to *make* those pages? NO!

Beyond that I apologize if I'm not "professional" enough for this forum. Of course, I'm not a professional pano-maker. I just design & build things for a living.

Last edited by touristguy87 (2009-02-24 22:42:58)

touristguy87 wrote:

If they just want a high-quality 700MP pano, that's quite another. You can't ignore the fact that you aren't actually taking a 700MP image with your camera, and that smartblend is  *creating* a 700MP pano for you from much-lower resolution sources, are you telling them this?

As is stated SEVERAL TIMES - and what you obviously still don´t understand:

Too print a photography in a given size you have to have enough pixels to do that.
Printig always means pixels/size.
Let´s say you wanna print A2 as offset (300dpi).
That means 59,2x42cm@300dpi
Let´s say you wanna print 7x2,50m@150dpi
That means 700x250cm@150dpi=41339x14764px=610,3MPx=1,71GB

So you have to shoot "big" enough to get a file of that size finally.

I did such a job shooting 6x17cm on 50ASA Velvia and drumscanned the tranny @6000dpi.
Another job needed to be printed at 20m wide @55dpi - which means a file of 433MPx and another needed 700MPx.

The point is to shoot so much pictures that a stitch of them all results in the final resolution WITHOUT having to be scaled up.

Using a precise head and a very good lens - in my case a manual Nikon 85mm - results in single pictures which each at 100% shows the final resolution resp. more in case of scaling down because having shot more as needed. Which is realistic to do for selecting the final proportions.
So if you need 500MPx better shoot 700MPx to have enough space to choose the final picture´s proportions.

Again: rendering a 700MPx pano is no problem at all - it´s the editing of such a big stitch what´s extremely time-critical.
Because you never use any editor - as you stated - you obviously can´t even imagine these facts.

Upscaling means clearly visible losses of details. Especially being printed big - you never know the final viewing distances. The 7m prints had to be sharp and detailled visible at 1,50m. That´s pretty close for such a big picture - but the client demanded it this way.
Anyway: making a photography and print for home use you bet people come very close to view it . . . . would you like to present pixel-mud to them? I don´t.

One of the advantages using a program like APP is the free choice of final resolution by choosing the number of shots and the lens to shoot them.
Shooting and stitching 150 pictures instead of one results in a resolution you would need an 8x10" film drumscanned @2000 or 4000 dpi (which i already did often) . .
But you would have big problems and also big costs to do it bracketed with enough steps to get a smooth exposed image in the end.
Very expensive, very time-consumpting, very heavy gear to carry, and some other things.

Believe me: it´s much easier to shoot those jobs digital using APP.

Besides: i told you that several times in at least three threads . . don´t you read or what?

"You can't ignore the fact that you aren't actually taking a 700MP image with your camera"

Oh yes - i am!! I just do it shooting, say, 200 single pieces of that image and stitch them together . . . . which in fact is the reason to use a stitcher (besides of shooting spheres for interactive panos).

Last edited by klausesser (2009-02-24 23:19:15)

"And it doesn't "create extra pixels out of nothing".

Taking a closer look at it: of course it does - where should the additional information come from?

"An interpolation procedure uses a mathematical transformation on the existing pixel data."

yes - what does that mean? You scale the existing information. In a pixel-based image that means you scale up the pixels . . by adding synthesized ones which didn´t exist before.
Scaling existent information means to blurr it. That means you blurr the image.

Can you make 10L of beer from 5L? No. You´d have to add some liquid to add volume.
No way you have the same beer afterwards.

"Again: rendering a 700MPx pano is no problem at all - it´s the editing of such a big stitch what´s extremely time-critical.
Because you never use any editor - as you stated - you obviously can´t even imagine these facts."

LOL I can't imagine *never* using the editor.

"Upscaling means clearly visible losses of details. Especially being printed big - you never know the final viewing distances. The 7m prints had to be sharp and detailled visible at 1,50m. That´s pretty close for such a big picture - but the client demanded it this way."

Sure. Now define "sharp and detailed".

""You can't ignore the fact that you aren't actually taking a 700MP image with your camera"

>Oh yes - i am!! I just do it shooting, say, 200 single pieces of that image and stitch them together . . . . which in fact is the reason to use a stitcher (besides of shooting spheres for interactive panos)."

No, you're not. You simply are not shooting 700MP images, you are not creating 700MP scans from your film negatives or whatever. Smartblend is an interpolation process just as much as anything else is.

>"Where did I ever say that everyone else is doing it wrong?"

"constantly."

It's easy enough to see that that isn't true just my reading my posts, isn't it?
I can't recall a single time that I have ever said that anyone else is doing it wrong much less "everyone".


"Can you make 10L of beer from 5L? No. You´d have to add some liquid to add volume.
No way you have the same beer afterwards."

No but you're not making beer out of nothing, either. You're really wasting time with this.
What you are saying is that 10L of beer which is undiluted, 100% "beer" is better than anything that you could make that is also 10L, half of which is undiluted beer.
That's simply untrue.

And it's likewise very very "exaggerating" to talk to pedants who refuse to think logically and who cling to their pedantic beliefs in the face of cold hard facts and good logic.

Last edited by touristguy87 (2009-02-25 00:01:00)

""An interpolation procedure uses a mathematical transformation on the existing pixel data."

yes - what does that mean? You scale the existing information. In a pixel-based image that means you scale up the pixels . . by adding synthesized ones which didn´t exist before.
Scaling existent information means to blurr it. That means you blurr the image. "

You're leaping from error to error to error here.
Interpolation is not merely "scaling up the pixels".
Scaling up a pixel means to increase its size.

The interpolation process redigitizes the image mathematically using the original data PLUS the algorithm and desired final dimension as the parameters.
Without knowing the algorithm AND the data all you can do is guess at what it is doing and what is going to come out of it.

But in any case Smartblend is doing the same thing. The difference in using your source data with smartblend and rendering to 100% is that smartblend doesn't have to change the spatial parameters of the whole image. It merely needs to digitize it. It does the same thing: makes use of an algorithm and the original source data to digitize the final pano.

If I do this by creating the pano at a lower resolution and then interpolating to get it up to the resolution that I want, the resulting IQ depends on the algorithm, the original data, the pano, and the final image parameters. IE it depends on the smartblend output plus the 2nd interpolation algorithm. IE it depends on a cascade of two interpolation algorithms applied to the original image data. Any difference in the end depends on how you qualify the result. Qualifying the result *scientifically* inherently depends on quantification of the final image, all of which depends on the source data and the 2 algorithms used. All you are doing by rendering the pano to 100% instead of rendering to 25% and then upsampling by 400% is avoiding the 2nd algorithm. You can still evaluate the result the same way. But you aren't evaluating it in ANY way other than DPI and the specific process used to generate the result.

So you are abstractly assigning "quality-points" to both the DPI and to the process of creation. You have no concrete evidence on which to base your statement that the cascade approach will not generate an image of acceptable quality.

In any case you should just try it with BlowUP! and/or Qimage and see the difference.
You can scale the whole issue down to a manageable size and print it out at some appropriate DPI and see what you get.
There's no reason to not try that. It will take 1.3x as long to render it twice, using smartblend (once at 100% and again at 25% or just downsample the 100% render to 25% and then upsample it with either of the aforementioned tools) as if you just rendered the pano at 100%. Then print both at a scaled DPI and examine the print with a magnifying-glass if you want.

If you're rendering 700MP to 55dpi for a 20m print, then print it at 1.5m at 600dpi. You want it on a smaller print, print a proportional section of the total image at 600dpi. You can then preview the final result at scale. It should be trivial to use the crop tool in your favorite image-editor to do this for enough sections of the print to give you a good idea of the final IQ especially if you have a good printer that can print much higher than 600dpi. Of course in this final step you have to remember that your printer is not just taking bitmap data from your computer, it's run *again* through an interpolation algorithm, which is the whole point of Qimage, to give you the highest interpolation quality possible when printing. But if you set it up correctly then you won't have to worry about the printer-driver interpolating the image again.

But you would get the same CoC with a 2m print at 500 dpi viewed at 10cm that you would get with a 20m print at 50dpi viewed at 1m.
So experiment.

Last edited by touristguy87 (2009-02-25 00:34:01)

ps if your issue is with the editor, I can't help you with that. My main issue with it isn't the speed (though that is an issue), it's that I see little benefit in mucking-around with it simply because it doesn't give me the information that I need to do what I want to do with it. I don't need to play with the control-points and there's not a lot of benefit in changing the other parameters. Some but not much. If I wanted a "perfect" pano I couldn't get one with APP anyway.

"And it doesn't "create extra pixels out of nothing".

Taking a closer look at it: of course it does - where should the additional information come from?

"An interpolation procedure uses a mathematical transformation on the existing pixel data."

yes - what does that mean? You scale the existing information.

I can't read? Can I quote?


"In a pixel-based image that means you scale up the pixels . . by adding synthesized ones which didn´t exist before.
Scaling existent information means to blurr it. That means you blurr the image. "

...this is a series of errors, as I said. You are, at best, talking about a comparison between one image which has been upsampled to a certain resolution to an original shot *at* that resolution.But you do not have an original *pano* at that resolution. You have a series of shots at that resolution which are then interpolated into a pano by APP.

You cannot avoid interpolation this way.

or digitally:

You´d NEVER get the details/sharpness by upscaling - believe me: i did explore it very intensively.

Last edited by klausesser (2009-02-25 00:47:48)

analog - 6x17cm EPR drumscanned @2000dpi

look ok let me spell it out for you. (I think that you know that I'm right and you're avoiding the issue as best you can)

Take one of your high-resolution images.
1-downsample it to 25%.
2-upsample it using blowup! (just for simplicity)

Now you have two images of the same total resolution in MP.

Take the same crop from each.

Post the originals and the crops (if you can)

tell us the image parameters

and let us see what happens

The full images that you post should look EXACTLY the same when viewed at full-image.

The CROPS might look different...

but if we know the crop-ratio then we can step back away from the monitor to see the crops in correct perspective.

Last edited by touristguy87 (2009-02-25 01:03:19)

"cool I realized you didn´t know what "resolution" means. If you want more examples: ask.

"I rest my case until either of you scare up the cojones to do what I asked and prove me wrong."

poor boy - asking silly questions. Having silly thoughts."

"How to avoid the issue" 101

"The full images that you post should look EXACTLY the same when viewed at full-image.

The CROPS might look different...

but if we know the crop-ratio then we can step back away from the monitor to see the crops in correct perspective."

As i said: the "crop ratio" is @100% - no matter how big or small the cut-out is: it shows the 1:1 pixels.

If you have a display of at least 3m width i send you a full picture @100% - no problem for me.

You also can order a print 3m wide printed @150dpi - also no problem.

I realize again that you simply didn´t get the relations of pixels and sizes and shooting high-rez and displaying it.

But - to be honest: it really IS a bit complicated. So i understand your problems to understand it . . we all had to start gathering knowledge.

DrSlony wrote:

touristguy87: and I'm waiting for your blown-up version of my 25% image, the one I posted above.

I'm not going to waste time blowing up your 25% downsample of a 600x600 source, dude.

Admit you're wrong or go away.