 |
|
|
|
|
|
 |
| ||||||||
 |
 
|
|
|
|
|
|
|
|
|
| ||||||||
|
|
| 
| User list | You are not logged in.
I often whine about this, so heres a quick example of why its better to tonemap or exposure blend AFTER stitching, not before.
AFAIK, tonemapping uses global light values, so theoretically when you do it on a whole panorama after stitching the result should be better than when doing it on the source images before stitching. On the example sunrise panoramas below you can see why - the uglier one was tufused before stitching which didnt work out too well. After tufusing, the photo that covered the sun turned out too muddy, some of the photos that covered the trees turned out bad because the edges of the trees were very dark while the centers were almost white (i had to manually pay special attention to these shots during tufusing and despite this I'm still not happy with the result of the first-tufused-then-stitched pano). Sure you could manually go through all source images and tonemap/exposure blend them individually to make sure you dont do "shit in = shit out", but that does take a long time, sometimes it just wont work, other times the result will still be different from what you'd achieve if you tonemapped/exposure blended the whole pano's layers after stitching. Different doesnt mean worse, and it sometimes does happen that a 'locally tonemapped' pano (by locally i mean the source images) appears to be more 'magical' than a globally (after stitching) done one. On these two example shots, the corrugated steel roof of the locally tufused one has more color variation, which might appeal to some people. I prefer the way the two main balconies and surrounding walls on the locally tufused pano look than on the globally done one. The trees are also more alive on the locally tufused pano, although too much for my liking. I could achieve a very similar and probably better effect on the globally tonemapped version if I'd add some local contrast, dodge, burn and tweak colours and saturation.
Which one do you prefer, why, and what are your thoughts?
Last edited by DrSlony (2008-08-01 06:50:36)
Online
Hi..
I prefered the first one because of the "normal" tones. The second has much more the HDR signature. In fact it depends on what kind of feeling you want to show in your picture. Bur for me it's the first, the best.
Offline
Yeah, I'd go with the first one any day... If the first is done with APP first and TuFuse second, that's the way I'd procede too.
Offline
I much prefer the first. The only thing I quite like about the second is the light coming down the road, as though lit by the rising sun. The trees look wrong. They look lit from the front whereas the light source is behind them.
Mike.
Offline
there is a talk about a new HDRI program over at www.hdrlabs.com which looks to give a better output then Photomatix...or maybe it is just my untrained hands that are lacking the proper skills to master or tame Photomatix :-)
Henrik
Offline
Both versions have a sky halo, which is one of the things I dislike about conventional tone mapping. Does Tufuse typically produce halos? Enfuse does not, but it can't handle large images on my PC.
Offline
Tufuse gives more realistic results than tonemapping. Locally tufused pano looks more like a tonemapped hdr, globally tufused pano looks more natural (first one is global, second local). You can also control the free tufuse with flags to minimize halos (I didnt, i used default settings) or buy tufusepro which has some added features over the freeware tufuse.exe and has a gui with a preview so you can better control fusion and reduce halos to your liking, quite nice! Both are windows-only but both work in wine.
ps. Woohoo! When you google 'tufusepro' my app forum post comes up #1 :]
http://www.tawbaware.com/tufusepro.htm
http://www.autopano.net/forum/t3562-tufusepro
Last edited by DrSlony (2008-08-01 21:10:43)
Online
I agree with Dr Slony that the best solution consist to fuse differently exposed layered panoramas than bracketed source images. The additional reason is that I'm frequently using a monopod and my source images doesn't correspond perfectly.
In my opinion, Enfuse or TuFuse gives better, more realistic results than Photomatix or FDRTools I was using before.
I made some interior panos of mountain chapels. You may find an exemple of Tufuse Pro blending on my site :http://annecypanoramique.free.fr/2008-0 … illon.html
and another one in attached file:
This one was shoot on monopod with NN3, 350D and Sigma 8mm at f=3,5, 400 ASA, 1/8s + 1/30s + 1/125s
Hub
Offline
I didnt knew that the size of attached file is limited...
Here you are a link to the second pano I mentionned above:
http://annecypanoramique.free.fr/Entremont_chapel.html
Hub
Offline
ad. Warren Sarle: also note that the halo effect is much more visible on the thumbnail than on the full-sized image.
Below is another test, global tufuse but with the pyramid levels parameter changed.
img number - pyramid levels (-l #)
1- 5
2- 6
3- 7
4- 8
5- 9
6- 10
Online
Last image, I promise :]
This is a level10 (l10 is the default level) with some postprocessing. I think a level9 with postprocessing would look best, but I did this l10 image before the above tests and now I'm too tired and too hungry to bother redoing postprocessing for l9 :]
Online
Well done DR Slony… reality is not revealing truth, never, but the feeling of our perceptions…
Our eye is constantly refocusing, readjusting levels… It's a magic tool that we can't control.
Our eyes are controlling our brain and gives us a wide range of FOV, focusing and adjusting light level…
We better start with our weak reality than try to overcome it 
This is not of course an argument but a base in intelligent dialogue… Of course !!!!
Offline
Reality is subjective, and I feel I'm cheating people when I show them a photograph and I claim that the scene really did look like that, since I know that a camera will never capture an image the way I saw it while shooting it, never, no matter what hardware, software and techniques we use! So having come to grips that reality basically ends at the end of my nose, I now feel much more comfortable to 'fix reality', makes photos look even better than what I saw while shooting. Cheat, but cheat well!
Online
To Slony
I cant blame…
Soft cheating seems to the the well acceptable to main human eyes… Isn't it !!
I do respect all approach in picture tokken when the 'general pubic' agree… even if its a new kind of style, like, pointillism was not understood in it's time… Claude Monnet was a good example when Gauguin and Paul Cézanne where lest interesting ones… too many people told me that those artist were not even able to draw properly… again very subjective feelings… halas
Offline
DrSlony wrote:
ps. Woohoo! When you google 'tufusepro' my app forum post comes up #1 :]
I suspect Google giving more weight to one's own posts but I'm not sure.
Questions:
- which bracketing (or only underexposure + RAW processing) ?
- which metering method (i.e. same exposure, matrix, histogram to the right) ?
- which APP mode (NONE, AUTO, anchors) ?
2 x Shooting options x Tufuse options x APP options = a rather large amount of possibilities !
|
Tufuse before / Tufuse after
Last edited by GURL (2008-08-02 16:27:39)
Offline
GURL hmm lets see:
- which bracketing - well that depends on the scene, if the horizon is quite flat then I can save time and memory card space by just exposing the ground and sky differently, but in this city scene where i have a low horizon and then houses shooting up from it and a sun rising between them bracketing is a must. I used -4, -2. 0, +2, +4. This isn't something that needs to be tried, this is more based on precalculation, the environment forces you to use the appropriate option!
- which metering mode - if its bracketed then it would be ideal if I could expose for the sky and let the camera automatically step down. Mine doesnt do that, so I either expose for the sky and manually step down -2 or -4 EV (2 if shooting 3 brackets, 4 if shooting 5 brackets), or just expose for the 'average' brightness object and hope for the best :] If its not bracketed, if its differently exposed layers well then do it normally as you would any photo (expose to the right and fix it in raw converter, since APP wont let me precisely control the exposure. I could also feed the exposed to the right raws into APP, render the 48bpp pano to 1 file per layer and merge them afterwards, but its best to decrease exposure for the exposed to the right photos in a dedicated raw converter)..
- which APP mode - hmm, I typically use LDR, twice I had to use none. Would be great if APP had curves! Then I could brighten it without blowing highlights. If I bracketed then in APP-1.4.x I do a little trick I might describe in another post if anyone is interested, it wont be necessary for APP-2.x. If layered then LDR, sometimes I use RH2.
- Tufuse options - well not many options here if we use the freeware version (and when I tested those above panos, the free version did better than my tweaked tufusepro beta! The preview looked nice but it came out bad) then I do -l9 and -l10, compare, discard uglier one. -w obviously, and -a 16.
- a rather large amount of possibilities ! - x2! 
- Tufuse before / Tufuse after - after if possible!
Well thats all my opinion, but I can back it up with many tests if anyone wants to "take part in a civilized polemic", ie. argue!
Online
I prefer the second image, the HDR tone mapped one. True, there are some HDR "artifacts," but the image is alive. The first image is flat and lifeless.
Offline
What about the last one, the one from 2008-08-01 23:34:08? i think its the best of both worlds! Its a global tufuse (I don't like the practice of calling tufused or tonemapped images HDR, they are not HDR and it gets confusing and unclear what someone is talking about).
Online
DrSlony wrote:
What about the last one, the one from 2008-08-01 23:34:08? i think its the best of both worlds! Its a global tufuse (I don't like the practice of calling tufused or tonemapped images HDR, they are not HDR and it gets confusing and unclear what someone is talking about).
Semantics aside, I would say that the last one is the best. Exposure blending or HDR processing is a choice in a program like Photomatix. This is just a tool. It's up to the artist to make it work.
Offline
DrSlony wrote:
Which one do you prefer, why, and what are your thoughts?
Well - the second picture shows the typical "beginners" faults using Photomatix without some experience.
My very own experience is to use Photomatix FIRST for generating .hdr and tonemapping it and save it as 16bit TIFF and THEN stitching the TIFFs in APP.
I tried all other ways using al the programs i icould get for Mac and PC and Linux.
Using Photomatix correctly produces great and very natural results - with some more fine-adjusting as, let´s say, Fuse.
Try "tone-compressing" instead of "tone-emhancing" - it works much more "neutral". But you can work very "neutral" or "natural" even using "tone-enhancement" if you know what you´re doing . . 
(which always is the "basic-problem" . . )
best, Klaus
Offline
360 Degree wrote:
Semantics aside, I would say that the last one is the best. Exposure blending or HDR processing is a choice in a program like Photomatix. This is just a tool. It's up to the artist to make it work.
You´re absolutely right! There is a lot of confusing when it comes to HDR - but there are very clear and straight terms.
.hdr is a 32bit/ch floating-point file - which can´t be displayed on monitors, displays or print.
It has to be "tone-mapped" down into a displayable format. While "tone-mapping" - "downsampling" to be displayable - you can make some mistakes . . . and get lollypop-coloured pictures which are referred to as "typical Photomatix" . . . 
In fact generating a "real" .hdr delivers the best source to make an extremely "natural" or "neutral" looking picture with a widespread tonal range that´s imposible to get from "normal" pictures.
Even "exposure-blendig" without "real" .hdr can´t generate such a range - but comes very close and works fine in most cases.
Using a plain .hdr only makes sense in using the 32bit/ch floating-point advantage which delivers a natural illumination for objects in 3D-scenes.
It´s then a combination of mirror/reflection-map - which reflects on glossy objects the scene that was shot as a sphere - AND an illumination BY the scene which was shot as a .hdr-sphere.
The shadows the object produces on a ground absolutely are the "correct" shadows which were present on the photographed situation.
So you can set a car into the sphere and it not only reflects the surrounding buildings or landscape - the car needs no additional lighting inside the 3D-scene because it´s lit by the .hdr´s ability as "global-lighting/ilumination"
This is the point using .hdr - an encreased tone- or exposure-range you also can get with "exposure-blending" and familiar techniques.
best, Klaus
Offline
Hi Klaus,
I think the "best" way depends on your audience. Most people prefer high contrast, high-saturation, bright colored, highly sharpened pictures, which is why P&S cameras do that. People who "know" photography prefer more natural pictures--so with a DSLR, you have to work in PS to get the results of a P&S.
And there are a few painting styles that prefer the "lollipop" colors of "bad" HDR.
Sometimes overuse of HDR is what is wanted--there are a few HDR images that look like "comic book pictures--the the best way I can put it is that colors are not especially bright, but colors are restricted to several blobs on the color wheel
You're doing the right thing for your customers, but other types of customers may prefer a little more color, and some digital artists may not care about reality, but want to show their vision.
Offline
DrSlony wrote:
- which bracketing - well that depends on the scene, if the horizon is quite flat then I can save time and memory card space by just exposing the ground and sky differently, but in this city scene where i have a low horizon and then houses shooting up from it and a sun rising between them bracketing is a must. I used -4, -2. 0, +2, +4. This isn't something that needs to be tried, this is more based on precalculation, the environment forces you to use the appropriate option!
APP is able to to produce well adjusted images from source images recorded using different exposure settings so that using -4, -2. 0, +2, +4 bracketing is something I would like to avoid for panoramas ! This is a wish, I don't know when this is possible or not...
Systematic bracketing looks as primitive as the methods photographers used before some sort of light metering was available ( when I was young I used the exposure rules for bright sun, hazy sun, cloudy weather, etc which were recommended for Kodachrome 25 film.)
When less than 5 shots are enough to ensure that:
- on the darkest one the noise level is decent everywhere
- on the brightest one none of the three color channels is saturated at any location
then the additional shots are useless (they just include a larger amount of noise or a larger amount of saturated false colors.)
I believe digital cameras still have enough hardware to guess which ones could be avoided (proof: if the light metering device can't meter the actual dynamic range of the subject the value the camera selects is just random !) Some sort of "intelligent bracketing" would be useful for non-panoramic shots as well, but I don't know how this could be explained to Mr Canon or Mr Nikon...
In most situations where I use a fisheye I examine the histogram or blinking shadows/highlights to under-expose to the right. When using a lens like a 28 mm equivalent one, where 30 shots are needed for 360 X 180 this is nearly impossible. For such a lens 5 levels of bracketing result in 5 x 30 = 150 shots!
Besides that, I have no doubt that after-fusing is better as my preferred method is to place the yellow anchor on the lowest exposure for a dark Smartblend version of the pano and to place the yellow anchor on the highest exposure for a bright Smartblend version of the pano and then to fuse them.
Offline
I agree that bracketing stuff on cameras is really a pain (no direct acces; you have to take a minimum of 3 shots where 2 are enough in 99%...): they have poor dynamic, and nothing to correct that! Intelligent bracketing should be really great, and really easy to do...
I will quickly work on papywizard to do such thing: running on a laptop (like a eeepc), it will possible to take a picture in remote mode, analyse it (histo. from jpeg is enough, I think), and take another shot if needed.
Back to this topic... I took a pano with the Merlin, with 3 shots at each position (bracketing). I first stitch them and generated 3 layers, then tufuse them: I got wrong positionning in some places... I then tried to first tufuse all shots, them stitch the result, and all was perfect. Why?
Offline
fma38 wrote:
I agree that bracketing stuff on cameras is really a pain (no direct acces; you have to take a minimum of 3 shots where 2 are enough in 99%...): they have poor dynamic, and nothing to correct that! Intelligent bracketing should be really great, and really easy to do...
I will quickly work on papywizard to do such thing: running on a laptop (like a eeepc), it will possible to take a picture in remote mode, analyse it (histo. from jpeg is enough, I think), and take another shot if needed.
Back to this topic... I took a pano with the Merlin, with 3 shots at each position (bracketing). I first stitch them and generated 3 layers, then tufuse them: I got wrong positionning in some places... I then tried to first tufuse all shots, them stitch the result, and all was perfect. Why?
Hi Frédéric!
Nearly in all cases i use Photomatix first and stitch afterwards. That works very fine. When i HAVE TO render hdr-spheres i stitch ALL bracketed shots first and render them as seperate bracketed layers for generating a hdr-file in Photomatix.
I NEVER had wrong positioning of the 3 panos when i generate a hdr-file in Photomatix - never needed to use the "align pictures" function. This works very well WHEN you need to align missmatched pictures.
best, Klaus
btw.: i tried to follow your postings in the french "Merlin" thread - but i can only guess what´s going on . . so sad . .
Last week i had an assignment i could have used the head very well - but my air-cable didn´t arrive in time and i couldn´t use the camera tethered. But i´m happy to have the head and your program!! Both are great stuff!!
Offline
