Tim Eastman
Posts: 47
Location:
Registered: 15 Nov 2006
|
Compression
Posted: 6 Apr 2008 at 0:46 GMT
|
reply
|
Would anyone be so kind as to share their compression techniques. With my D80 and 10.5 lens, shooting RAW > TIFF images stitched with PTGui and converted in Pano2QTVR at a 60 compression setting are anywhere from 1.25 to 3 meg files. I have seem some beautiful full screen panos which must be smaller than 1.25 meg. Any ideas? Thanks.
|
|
alert moderator
|
|
spyboy
Posts: 161
Location: Manchester, NH, United States
Registered: 7 Oct 2006
|
Re: Compression
Posted: 6 Apr 2008 at 1:19 GMT
|
reply
|
|
what is your base .jpg size to start with? I use 6000x3000 but my .swf files are 4+ mb
|
|
alert moderator
|
|
Tim Eastman
Posts: 47
Location:
Registered: 15 Nov 2006
|
Re: Compression
Posted: 6 Apr 2008 at 1:24 GMT
|
reply
|
|
6000 X 3000 also. I output .mov files as I have not explored the various alternatives.
|
|
alert moderator
|
|
michael medina
Posts: 160
Location: United States
Registered: 27 Jan 2008
|
Re: Compression
Posted: 6 Apr 2008 at 2:38 GMT updated: 6 Apr 2008 at 2:41 GMT
|
reply
|
file size depends on the level of detail and colors (something with green images taking up more room)
i just added a full site's worth of 1352 cube faces (originally shot at 4100 x 8200) compressed at jpeg medium (50%) and they are generally 1.25 mb to 1.9 mb. the file sizes are listed under the preview images so the user can make an informed click. pdxvr.com
sharpening increases detail and thus larger file sizes.
the smallest movie on my site is .7mb pdxvr.com/vr-6-theatre-lores but it's 1/4 totally black carpet.
the largest movie is pdxvr.webhop.org/vr-18-2 at a whopping 8.3mb which i am sure is due to all the pea gravel detail (just add "-lores" to the end if you can't wait for 8.3mb)
fullscreen obviously depends on your screen size too. all my vr's look fine (imo) at 1280 x 800, but go bigger and the low res start to get soft.
compressing at 40% and lower is possible in some cases, but generally if a file wants to be large, you have to let be large (or crappy )
|
|
alert moderator
|
|
michael medina
Posts: 160
Location: United States
Registered: 27 Jan 2008
|
Re: Compression
Posted: 6 Apr 2008 at 8:59 GMT updated: 6 Apr 2008 at 9:04 GMT
|
reply
|
heh i totally posted the wrong url to the large file there
the correct is pdxvr.com/vr-18-2
that other one is on my machine which is really slow  i've redirected it because i saw someone looked at it, my apologies to that person who waited for that download
|
|
alert moderator
|
|
mhc1
Posts: 112
Location: Hamburg, Germany
Registered: 29 Dec 2005
|
Re: Compression
Posted: 6 Apr 2008 at 21:15 GMT
|
reply
|
Mike: I am using 8000 x 4000 TIF shooting: RAW, converting with Capture One V4 - no further sharpening etc, just a little highlight/shadow control.
Output PTGUI ab. is 8000 x 4000 TIF 16 Bit (Autopano has slightly more)
What would you suggest ? At the moment I use this TIF - file with Pano2QTVR, settings: ............... Tile size: 1024 px Subdiv: 1x1 Quality: 76
These don´t effect quality I guess: Window size 1240 * 1000 Preview track enabled Resolution 1:8 Greyscale
Optimization: selected ( ... QVTR Flattener) Quality LOW for motion MAX for static
Gamma : 1.0 Interpolator: spline 36 ........................
I want my panos to become a little more sharp though. Using UltrasharpenPro ist OK but my TIF is much much sharper then the generated MOV. Also my JPG at 3071 * 10536 lack a of sharpness.
Any suggestions ?
PS: I don´t want to capture this thread , but is looks suitable 
ciao Mike www.360ded.de
|
|
alert moderator
|
|
Hans Nyberg
Posts: 620
Location: Denmark
Registered: 28 Aug 2005
|
Re: Compression
Posted: 6 Apr 2008 at 21:28 GMT updated: 6 Apr 2008 at 21:30 GMT
|
reply
|
michael medina said: file size depends on the level of detail and colors (something with green images taking up more room)
i just added a full site's worth of 1352 cube faces (originally shot at 4100 x 8200) compressed at jpeg medium (50%) and they are generally 1.25 mb to 1.9 mb. the file sizes are listed under the preview images so the user can make an informed click. pdxvr.com
)
I assume that youy use Mac when you are talking about 50% medium.
Just so that Tim who uses Pano2QTVR does not believe that 50% is a useable compression for quality with Pano2QTVR , Pano2VR and a couple of other Win applications.
Just for info 50% PhotoshopWEB Win/Mac and 50% in all other Mac applications applications i know of is compareable to 70% Pano2QTVR.
Hans
|
|
alert moderator
|
|
michael medina
Posts: 160
Location: United States
Registered: 27 Jan 2008
|
Re: Compression
Posted: 6 Apr 2008 at 21:50 GMT
|
reply
|
sometimes i wonder what i'm talking about anyway 
yeah medium in cubic converter and 50% in photoshop save for web. i'm not familiar with the products listed, so i should have mentioned that.
thanks for the clarification Hans
|
|
alert moderator
|
|
gus
Posts: 308
Location: United Kingdom
Registered: 19 Jun 2007
|
Re: Compression
Posted: 6 Apr 2008 at 22:34 GMT updated: 6 Apr 2008 at 22:39 GMT
|
reply
|
The tile size is directly related to the FOV. Having a tile size any bigger than the "optimal" value by p2qtvr will be wasting pixels. Only problem (bug imo) is that p2qtvr uses the "initial FOV" to calculate the "optimal" FOV, instead of the min-FOV.
For example, on my settings page, I have chosen a min-FOV of 50, and an initial FOV of 70. P2qtvr suggests an optimal tiles size of 857. The problem is, this size is based on the initial-FOV of 70, and if you accept this tile size, pixelation will be visible as soon as you zoom in beyond the initial FOV, which is obviously not ideal. A trick to force p2qtvr to use your min-fov instead of the intial FOV for its calculation, is to temporarily change the intial-fov (70 in my case) to the min_fov (50 in my case)on the settings page. Click outside the box, and p2qtvr will re-calculate the tile-size, which spits out 1287 for this example. Remember to change the values back!
So, since I am limiting my movie to a min-FOV of 50, there is no need to have the tile size any bigger than 1287.
What I normally do is in the preview window decide on the min-fov that I want my visitors to be able to zoom into (usually around 50-60), and then temporarily change the initial-fov in the settings box to get p2qtvr to calculate the tile size for me. That way, I know my tile size is no bigger than it needs to be, and that my movie will not shows pixelation at the min-fov.
Now that the tile size is accurately fixed, I can play around with the jpeg slider until I have a desired movie size.
gus
|
|
alert moderator
|
|
Smooth
Posts: 1440
Location: Australia
Registered: 21 Jul 2004
|
Re: Compression
Posted: 7 Apr 2008 at 4:56 GMT
|
reply
|
mhc1 said: I am using 8000 x 4000 TIF shooting: RAW, converting with Capture One V4 - no further sharpening etc, just a little highlight/shadow control.
When I last looked Capture One V4 or the "Pro?" version didn't deal with or have any option to correct Vignetting or Chromatic Aberration.
Is this still the case? If so, you don't even get to fix the BIGGEST issue with RAW fisheye images.
Regards, Smooth www.smooth360.info
|
|
alert moderator
|
|
mhc1
Posts: 112
Location: Hamburg, Germany
Registered: 29 Dec 2005
|
Re: Compression
Posted: 8 Apr 2008 at 6:52 GMT
|
reply
|
Smooth this is right. I use the method John mentiond before. The biggest issue of my RAW I think is the 14 Bit and the exposure (WB and highlites % shadows) CA´s are pretty good to reduce in PS. But can you tell me, if my workflow above is OK? How can I get more shrapnes into my pictures ?
I want them to be more crispy. Maybe I have to go for a Tokina 10-17 ???
Still didn´t buy it - I am not sure that this is a solution. Ciao Mike www.360de.de
|
|
alert moderator
|
|
Matt Rogers
[360 Precision]
Posts: 153
Location: Oxford UK, United Kingdom
Registered: 16 Jun 2005
|
Re: Compression
Posted: 8 Apr 2008 at 8:50 GMT
|
reply
|
The tile size calculation in p2qtvr is wrong though and really needs to be changed. The tile size should nothing to do with the FOV but the size of the equirectangular image.
The tile size can be calculated using the simple formula 1/pi x X where X is the width in pixels of the equirectangular image. So for a tile size of 1287 your ER image should be 4044 pixels wide. If you're using any other image size then you're losing more quality as the tiled images are being interpolated as well as transformed. You really don't want the source image to be interpolated more than once if you can help it.
To obtain optimal image quality you need to work backwards in the workflow to pre-determine the various image sizes/resolutions at each step.
I'll try and put together a sample workflow later this week.
Matt
|
|
alert moderator
|
|
Matt Rogers
[360 Precision]
Posts: 153
Location: Oxford UK, United Kingdom
Registered: 16 Jun 2005
|
Re: Compression
Posted: 8 Apr 2008 at 9:09 GMT
|
reply
|
mhc1 said: The biggest issue of my RAW I think is the 14 Bit and the exposure (WB and highlites % shadows) CA´s are pretty good to reduce in PS. Ciao Mike www.360de.de
What do you think 14bit has to do with your problems ?
mhc1 said: I want them to be more crispy. Maybe I have to go for a Tokina 10-17 ???
Still didn´t buy it - I am not sure that this is a solution. Ciao Mike www.360de.de
Buying even another lens won't help until you get your process sorted out. I see this all the time, customers upgrading camera and lenses to improve the quality and then end up disappointed. Even a D40x/8mm f3.5 can produce really sharp fullscreen (1600 x 1200) panos.
From your workflow above you're wasting all the resolution going from an 8000 x 4000 image down to a 1024 tile size, this is just crazy. You could have the best set-up in the world and it won't help if you do the above.
You need to understand what happens to your image at each step going from the Camera --> Raw Converter --> Stitching App --> Output App. Until you really understand this process all you're doing is guessing and your panoramas will never be optimal.
I learnt all of this before internet forums and email lists the hard way, through trail and error. You can post as many questions as you like all over the internet but you need to understand the underlying processes and concepts. Otherwise every time you get stumped you end up relying on other people to solve your problems.
Matt
|
|
alert moderator
|
|
Hans Nyberg
Posts: 620
Location: Denmark
Registered: 28 Aug 2005
|
Re: Compression
Posted: 8 Apr 2008 at 9:10 GMT
|
reply
|
Yes Matt is right. Both Pano2QTVR and Pano2VR are a mess. For people who do not know how to do it they are very misleading.
And the setup for conversions to cubefaces and between different projections has an interface which is not working.
Hans
|
|
alert moderator
|
|
Tim Eastman
Posts: 47
Location:
Registered: 15 Nov 2006
|
Re: Compression
Posted: 8 Apr 2008 at 15:47 GMT
|
reply
|
|
Regarding Pano2QTVR, it is easy enough to generate cube faces, but putting them back to equi is problematic. I can drag them onto the droplet, but I have trouble doing the same thing inside the program. Is this normal?
|
|
alert moderator
|
|