Mutli-resolution panos designed only for display in krpano or similar  

Got some great idea or a feature request? Post it here and discuss it. The most requested concepts are usually implemented, as Autopano Pro / Giga is very community driven.
no avatar
bradtem
Member
 
Posts: 210
Joined: Tue Nov 18, 2008 11:42 pm

Mutli-resolution panos designed only for display in krpano or similar

by bradtem » Wed Sep 26, 2012 2:18 am

As we start building bigger and bigger panos, the question comes up about how to display them. Printing at more than 300 ppi is wasteful, so unless you are covering a wall you are not going to use the resolution of a 5-row in a printout on typical large format printers most of the time.

And I'm noting two things:

a) The only way people are going to see this pano at full resolution is inside krpano or perhaps the gigapan viewer
b) In many cases, a lot of the pano consists of things that don't need to be high resolution, such as sea and sky.

You can shoot multi resolution, where you do the general background or sphere with wider angle, and use long lenses for the detailed interesting part in the middle. In fact, sometimes you have to do that because APP and similar programs can't actually do blends in the sky or sea unless you are using a computerized pano head which records a file of where the shooting was.

But if you want to render to capture that full resolution you have a problem in that the result is huge, even if only a stripe of the pano is at the full res. You may also find it handy to shoot certain "points of interest" in a lower-res scene using a long lens, knowing this is what people are going to zoom in on in krpano, and that they'll never zoom in on boring stretches.

So I would like to see both APG and the viewers move towards the idea of a multi-res pano, where the pano is output as a series of tiles (as krpano likes to display) but not all the tiles are at the same depth and resolution. krpano is already almost there, of course, it would just need to understand that some areas of the pano allow zoom and others do not, and handle the borders.

So in each tile zone, the tiles would be rendered deep enough to show the full resolution in that zone, but no deeper.

If this were possible, many people would shoot panos where they do perhaps even a full sphere with a fish eye, but then do the horizontal row or rows at higher resolution, and further go and shoot the points of interest with even higher resolution, generating a pano that the user can wander through and which appears to be amazing res without being a huge, huge size that takes hours and hours to render.

This is no minor change of course, but I think it's likely to be the way to go for online presentation and virtual tours. Indeed, one might automatically generate tags to go into the viewer that say "here is a place you want to zoom in" so people browsing the pano know what to investigate. It's possible that it might be nice to have an interface where I apply those tags to my long-lens shots of the points of interest, so it's all automatic. (ie. I caption those shots, or can caption them in the layer editor.)

Of course, I also still like to print, so I would need a render mode that lets me render at a printing resolution.

You would need the krpano folks to do this too, but their changes are not too bad, and you seem to be friends.

I should note that step one of this is also quite useful -- which is a rendering mode which outputs tiles rather than (or in addition to) a full image. Or in its complete form, a rendering mode which outputs what krpano's tools do -- a set of tiles, and an xml file with information on them, and a preview. This actually has a number of advantages over the approach of building the whole pano and feeding it to krpanotools. It is much faster and more convenient of course, but also reduces transcoding, and has no size limit issues for very large panos (which run into the limits of jpeg and tiff.) Of course if you do tiles + full pano you still have those issues, but it's still faster and more convenient.
Last edited by bradtem on Wed Sep 26, 2012 2:29 am, edited 1 time in total.

User avatar
billjones
Member
 
Posts: 101
Joined: Sun Jul 05, 2009 7:38 pm
Location: Newberg Oregon

by billjones » Wed Oct 03, 2012 12:54 am

What's wrong with the way krpano handles it now? I have several gigapixel images that where processed by krpano, I also did them with zoomify, but prefer the krpano presentation. So since I see no issues, I don't understand what yours are. My gigapixel-krpano samples are at http://rhmpano.com under gigapixel on the sidebar. As for prints, I was in the gigapan offices a week ago and they have what must be a 7 foot by 12 foot print.

Gigaimages also make great tiled murals. http://bxbweb.com/exampleimages/allisoncouch30x45.jpg

no avatar
bradtem
Member
 
Posts: 210
Joined: Tue Nov 18, 2008 11:42 pm

by bradtem » Wed Oct 03, 2012 6:05 am

You may not be understanding what I am talking about. The panos I am describing have shots with different resolutions. APP renders such at the resolution of the highest res (longest lens) shot. If you ahve a pano made of say 50mm shots that is 1 gigapixel, for example, and you insert a 200mm shot or two of certain areas of interest, APP will generate a 16 gigapixel pano. This is immense, hard to work with (impossible in many tools) and highly wasteful.

In another example say you are shooting a skyline from a distance. You might wish to shoot the skyline at a long focal length and get maximum detail, but there is little to be gained by shooting the sky above (and the sea below if apprporiate) with the long lens. In fact, in many shooting style syou can't shoot the sky that way unless you have a motorized panohead, as APP can't find control points. So instead of having a pano with huge numbers of gigapixels if you could get it to blend you don't have a pano at all. But really you don't want to waste CPU or disk on rendering the sky and clouds at this resolution.

So in the appraoch I describe, in the 1st example, you would get a pano not much more than a gigapixel, and you would be able to zoom in on key points as though it were a 16 gigapixel pano.

This would be extremely useful, and would lead photographers to capture a lot more high-res panos. Today capturing a super-gigapixel is a time consuming and computer consuming task. Make it easier and more people will shoot them. And this is a benefit even to the best of us. Trying to shoot a large photo at a long focal length requires lots and lots of time, much of it wasted on things like sky or other terrain of low interest. During that time, things change -- the clouds move, the light changes and more. These changes result in poor blends and bad looking images.

I'm amazed that anybody would doubt the value of this approach. I print giant panos, but if you have a 44" high printer and 300 pixels/inch (which is about all you will do) you only need 13,200 pixels high. That's 3-4 rows high on the 5D2 for example. There is no reason to shoot giant mutli-rows much higher than this other than to crop or use a zooming viewer like kprano. As such I believe that programs like APP should start working towards support for this approach.

User avatar
billjones
Member
 
Posts: 101
Joined: Sun Jul 05, 2009 7:38 pm
Location: Newberg Oregon

by billjones » Wed Oct 03, 2012 6:33 am

Thank you, I didn't understand the multi focal length approach was what you where getting at. Even if you did say so :-)

no avatar
leifs
Member
 
Posts: 612
Joined: Sun Sep 06, 2009 12:08 pm
Location: Ørsta Norway

by leifs » Wed Oct 03, 2012 8:24 am

I understand and I agree very much.

leifs
Olympus OM-D E-M1, Panasonic 8mm f3.5 fisheye, Leica 25mm f1.4, Olympus 75mm f1.8, Canon 100-400mm f/4.5-5.6 L
Seitz VRdrive2
Intel i7 980X, 48GB RAM, Win7 64bit, SSD RAIDs

no avatar
klausesser
Member
 
Posts: 7883
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany

by klausesser » Wed Oct 03, 2012 1:29 pm

bradtem wrote:Printing at more than 300 ppi is wasteful, so unless you are covering a wall you are not going to use the resolution of a 5-row in a printout on typical large format printers most of the time.

Hi Brad!

Printing at 300dpi is only needed for offset-printing (news-magazins/books, advertising folders) or high-end Lightjet/Lambda for artwork (Gursky and so on)
You must distinguish the image-resolution (pixel) from the resolution the image gets printed resp, displayed (dots).
Correctly the image-resolution is ppi (pixel-per-inch) and the print-resolution is dpi (dots-per-inch).

bradtem wrote:So I would like to see both APG and the viewers move towards the idea of a multi-res pano, where the pano is output as a series of tiles (as krpano likes to display) but not all the tiles are at the same depth and resolution. krpano is already almost there, of course, it would just need to understand that some areas of the pano allow zoom and others do not, and handle the borders.

So in each tile zone, the tiles would be rendered deep enough to show the full resolution in that zone, but no deeper.

If this were possible, many people would shoot panos where they do perhaps even a full sphere with a fish eye, but then do the horizontal row or rows at higher resolution, and further go and shoot the points of interest with even higher resolution, generating a pano that the user can wander through and which appears to be amazing res without being a huge, huge size that takes hours and hours to render.

This is no minor change of course, but I think it's likely to be the way to go for online presentation and virtual tours. Indeed, one might automatically generate tags to go into the viewer that say "here is a place you want to zoom in" so people browsing the pano know what to investigate. It's possible that it might be nice to have an interface where I apply those tags to my long-lens shots of the points of interest, so it's all automatic. (ie. I caption those shots, or can caption them in the layer editor.)

Of course, I also still like to print, so I would need a render mode that lets me render at a printing resolution.

You would need the krpano folks to do this too, but their changes are not too bad, and you seem to be friends.

I should note that step one of this is also quite useful -- which is a rendering mode which outputs tiles rather than (or in addition to) a full image. Or in its complete form, a rendering mode which outputs what krpano's tools do -- a set of tiles, and an xml file with information on them, and a preview. This actually has a number of advantages over the approach of building the whole pano and feeding it to krpanotools. It is much faster and more convenient of course, but also reduces transcoding, and has no size limit issues for very large panos (which run into the limits of jpeg and tiff.) Of course if you do tiles + full pano you still have those issues, but it's still faster and more convenient.

There are ways already to use different resolutions in parts of an image - for sky and sea as example.

But to be honest: dealing with that i found it definitely much more inconvenient than shooting the whole thing using one an the same resolution. It was very hard to do it in a way not realizing zone-breaks when zooming in or printing big.

best, Klaus
Simplicity is the keynote of all true elegance. Coco Chanel

no avatar
bradtem
Member
 
Posts: 210
Joined: Tue Nov 18, 2008 11:42 pm

by bradtem » Wed Oct 03, 2012 5:04 pm

I disagree. Today's printers all work from pixels and do their own internal job of mapping that to dots. Their resolutions in dots range from 600 to 2400 dots per inch but the goal is to render ppi. And my point is, if you generate a panorama for printing, you do not gain anything havng that panorama be at more than 300ppi. In part this is simply because of the resolution of the human eye when brought close to the picture.

My whole point is that it is indeed to do the multi-res at present. I am requesting a core feature to make that work better, and you may be right that other areas of the program need to be improved to make tha thappen. Consider this gigapixel image of mine http://pic.templetons.com/cgi-bin/imget?f=brad/pano/sfba/big4.jpg&fw=49494 This was shot with a long lens on the city but there was no choice but to shoot the sky and sea at lower res because APP can't blend those on its own. This was made for printing -- it's the 11' long back wall of a bar in San Francisco -- but I also like to show the full res in krpano.

no avatar
klausesser
Member
 
Posts: 7883
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany

by klausesser » Wed Oct 03, 2012 6:40 pm

bradtem wrote:I disagree.

No problem - you didn´t get me:

bradtem wrote:And my point is, if you generate a panorama for printing, you do not gain anything havng that panorama be at more than 300ppi.

That´s what i said: you don´t even need 300dpi at all - you must regard the image-resolution related to the output-size and this one related to the way of printing.

A print-resolution always is connected with a print-size. Big prints are - usually - viewed at a larger distances. So the image-resolution can be around 50-150dpi - depends on each single case.

I had a print done at 20mx4m using 50ppi image-resolution for outdoors.
What they used as printer-resolution i don´t know. The result was extraordinarily good - usually this sizes are printed at lower resolutions.

Another case was a 7m width print indoors - which means rather close viewing distance related to the size. The client wanted an image-resolution of 150dpi@7m width. The result was: you can sit beside the print in 1m distance and it looks very fine.
i used a 6x17cm transparency and scanned it @4000dpi on a drum-scanner (couldn´t use stitching because of fast moving objects).

300dpi image-resolution @print-size is only needed A) for offset - magazines, advertising folders or so or B) for fine art printing using Lambda or Lightjet as examples (Gursky and so). These machines have a fixed machine-resolution of 300 resp. 400dpi . . and fine-art prints - no matter off their size - often are viewed at rather close distance (i use to look at a Gursky very close - it´s great most of the time).

But even here the image usually is about 200dpi@print-size.

So: when you want to print 3mx4m as example you unlikely need more than 80-120dpi as image-resolution for your rendering. BUT: that also depends on A) the viewing distance resp. the kind your print will be presented and B) the kind of print-technology which will be used in combination with the print-material . . canvas for example needs lesser dpi than paper. Not surprising at all regarding the surface-structure. . . . :cool:

best, Klaus
Last edited by klausesser on Wed Oct 03, 2012 6:48 pm, edited 1 time in total.
Simplicity is the keynote of all true elegance. Coco Chanel

no avatar
bradtem
Member
 
Posts: 210
Joined: Tue Nov 18, 2008 11:42 pm

by bradtem » Thu Oct 04, 2012 12:32 am

We're still at cross purposes. DPI is an almost irrelevant number these days. It only matters to the people who design how printers will do their dot patterns -- it should not generally be used by people doing colour prints, which are today all generated from pixel based digital images and PPI is the important number. A lot of people confuse them, and will say DPI when they really mean PPI. That's because long ago in the days of certain B&W prints, in particular text, the DPI actually was a number you cared about. Sadly, how many PPI a printer can do given its DPI is not a fixed ratio, it depends on the printing technique, droplet size, accuracy etc. Many people are satisfied with 1200dpi printers to generate 300ppi, some are satisfied with 600dpi, and some demand even more than 1200dpi.

So let's not use DPI if possible unless we are discussing inkjet printer design.

At any rate, while normally you are right that large prints are not viewed close up, so you can print them at 50ppi or worse, the whole point of gigapixel photography is to produce prints which, unlike traditional photos, can be printed at 200 to 300ppi and thus are tack sharp even if you put your nose up to them, but at the same time are giant. That's what is special about these photos in printed form. People who do giagpixel panos are not interested so much in low-res printing.

As I said, I was making something to fill a wall, 11 feet high, and still be sharp up close, and a gigapixel shot is what you want. But I had to shoot it as two panos -- a 200mm shot over the central party with the city, and a few rows of 50mm shots with the sea and sky. The sky at least looks the same at lower res as it does at high res, or close enough.

But such printings are rare. I shoot a lot of large panos and only have had 3 or 4 of them printed 10' high. The vast majority are only seen at their full resolution in a zoom viewer, and I think it's the same for many other folks here, so a way to make them without them being files that bring machines to a crawl, and without having to shoot long stretches of boring stuff like sky that won't find control points --- that's a huge plus. When we want to print them, then we can ask APP to render them more full size for printing. But for me, it will be:

95% of the time -- render them for full res zoom viewing, in which case the multi-level resolution is a big win
4% of the time -- render them about 7500 pixels high for printing on my 2' wide roll printer
1% of the time -- render them at full resolution for printing wall-high.

Because of those ratios, it's clear that a focus on zoom viewing is valuable, at least to me.

no avatar
klausesser
Member
 
Posts: 7883
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany

by klausesser » Thu Oct 04, 2012 1:51 pm

bradtem wrote:DPI is an almost irrelevant number these days.

No - when it comes to print an image it definitely IS relevant . . :cool:. How would you get an image on paper without printing dots on the paper? :D

Lets make it clear and simple: having a digital image (or a scan) means to have a heap of pixels, right? Yes.

Displaying the pixels means converting them to points/dots for making the pixels visible on devices, paper, screen and so on. Right? Yes.

So the first thought shooting hires-panos is:

HOW much zoom do i need/want for using it on screen?
HOW big do i want to print the pano?

Depending on these decisions choose your lens/camera and the way you shoot.

An example:
When i need a sphere for moderately deep zooms i use to choose an 85mm lens. This gives me 86000 x 43000px for an equirectangular image.

This amount of pixels tells me how big i can print the image - which means dpi. By default APG is set to .72dpi. Setting it to let´s say 300dpi gives me a smaller amount of dpi for the print - but the amount of pixels stays the same.

This print still contains the 86000 x 43000 pixels . . . but the size of the print depends on what i set as dpi. Very simple.

Conclusion in the sense of photography (ANY photography): always use the best resolution you can afford to realize - which means sensorsize and focal-length to shoot with.
Downscaling an image always is possible without loosing quality (almost).
But upscaling an image definitely means a loosing quality - always.

Always render an image @100%. Always calculate before the shooting how much resolution you need in dpi for printing or displaying big.

If you want to print 20m x 4m you need to use a lens/camera which provides the amount of pixels you need for printing it at the needed size in dpi.

Basically it´s that simple: bringing pixels to paper or displaying them anyhow needs dpi/ppi . . ;):cool:

best, Klaus
Last edited by klausesser on Thu Oct 04, 2012 2:31 pm, edited 1 time in total.
Simplicity is the keynote of all true elegance. Coco Chanel

no avatar
klausesser
Member
 
Posts: 7883
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany

by klausesser » Thu Oct 04, 2012 2:28 pm

bradtem wrote:the whole point of gigapixel photography is to produce prints which, unlike traditional photos, can be printed at 200 to 300ppi and thus are tack sharp even if you put your nose up to them, but at the same time are giant. That's what is special about these photos in printed form. People who do giagpixel panos are not interested so much in low-res printing.

That´s not limited to stitched panos. I still use to use 8x10" transparencies sometimes for my regular jobs as an advertising photographer. Scanning them @2000 or @4000dpi on my drum-scanner provides some 4-6 gigapixels (depending on bit-depth) - from ONE SHOT.
That allows quite big printing.

But stitching basically has virtually no limits in terms of size of course . . :cool: Limits rather exist in the handling of data- and display/print-sizes.

We recently presented a pano in a cinema with digital projection on a real big screen - that was phantastic!

Basically it´s - as always - a question of clever balancing in- and output. :cool:

best, Klaus
Simplicity is the keynote of all true elegance. Coco Chanel

no avatar
bradtem
Member
 
Posts: 210
Joined: Tue Nov 18, 2008 11:42 pm

by bradtem » Thu Oct 04, 2012 5:42 pm

Ah, I must admit I have not had the opportunity to print a 8x10 transparency at full resolution for projection. I agree that is an interesting application.

But alas, I must continue to disagree about DPI and PPI. PPI is again what you are always interested in. Inkjet printers render images by spraying tiny dots, dots that are much smaller than a pixel, which is why you might use a 1200dpi or 2400dpi inkjet printer to output your 300ppi picture. Some printers, like a lightjet, emit a single pixel as a single multi-coloured dot, and the dpi and the ppi are the same. Much of this confusion comes because in the old printing days, DPI was the only number people used, and it originated from using dot matrix printers and laser printers, and to some extent halftone screens to print analog photos (though there they would measure the screen in lines per inch though it was lines of dots.) Analog photos don't have pixels, and their resolution, particularly in b&w, is very high though not infinite.

But then digital imagery was born and was all about pixels. The confusion is that pixels are dots (which can be any colour) but they are not related in most case to printer dots (which can be of only a few colours but can vary in size and are usually much to small for the eye to resolve.)

The confusion has grown further because many software tools -- including the venerable photoshop which probably drives the rest -- incorrectly express picture resolution for printing with the term DPI rather than PPI. DPI is an older term and is around for legacy reasons, but it is generally incorrect, and I suspect the folks at Adobe and other companies know this but don't want to change it. The truth is that a large fraction of graphic designers use DPI when they mean PPI. The main issue is the confusion it causes when printer manufacturers take advantage of the fact that their printer is 2400DPI (in real dpi) and because so many people think that DPI is PPI, this lets them advertise the printer as having a ridiculously high resolution.

As noted there are situations where you can argue that a dot and a pixel will be the same, and thus DPI == PPI. A scanner is an example, though since the output of scanners is pixels I think it's wiser to talk about scanning at 2000 PPI and not 2000 DPI. And If you are printing a binary image (sometimes called line drawing mode) an inkjet or laser printer can have DPI == PPI. But today, continuous tone printing (where dots are pixels) such as on the lightjet or a film recorder is rare, and almost all printing these days is halftone, where dots and pixels are very different things.

User avatar
billjones
Member
 
Posts: 101
Joined: Sun Jul 05, 2009 7:38 pm
Location: Newberg Oregon

by billjones » Thu Oct 04, 2012 6:56 pm

What was the topic???

no avatar
klausesser
Member
 
Posts: 7883
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany

by klausesser » Thu Oct 04, 2012 7:54 pm

bradtem wrote:But today, continuous tone printing (where dots are pixels) such as on the lightjet or a film recorder is rare, and almost all printing these days is halftone, where dots and pixels are very different things.

Most commercial printing is done in offset - news-magazins, folders, leaflets, book-covers and so on. Ink-printers cover another part of the market. Digital C-print like Lightjet/Labda is very expensive - but state-of-the art in the art-business.

As an advertising photographer i have to know what clients and their producers need to get a project done.

Discussing about whether it´s points or dots isn´t part of the business . . . :D - everybody in the business knows what´s up when it comes to pixels or dpi or ppi . . . :cool:

best, Klaus
Simplicity is the keynote of all true elegance. Coco Chanel

no avatar
bradtem
Member
 
Posts: 210
Joined: Tue Nov 18, 2008 11:42 pm

by bradtem » Fri Oct 05, 2012 5:17 pm

Bill, indeed the topic is really "what do you do with your giant pano once you make it." And today I'm finding that by far, the most common purpose is display in a zoom viewer like krpano, though I still print reasonably often on an ordinary large format, and occasionally sell to somebody who wants to print wall high.

As such, my goals for how my pano render works are changing, and are starting to orient towards effcient and high resolution viewing of the important parts of a pano in a program like krpano. Thus the suggestion of APP rendering as tiles for krpano, and of having it render different sections at different resolutions (deeper sets of tiles) and adapting krpano to deal with different depths in different sections.


Return to Autopano future

Who is online

Users browsing this forum: No registered users and 1 guest