Manfrotto 303SPH and Panasonic Lumix DMC-FZ8  

In the panorama field, hardware is also part of the success. You can discuss here about it: camera, computer, pano head, anything
User avatar
John_Sauter
Member
 
Topic author
Posts: 173
Likes: 0 post
Liked in: 0 post
Joined: Sat Jul 28, 2007 4:59 pm
Location: New Hampshire, USA
Info

Manfrotto 303SPH and Panasonic Lumix DMC-FZ8

by John_Sauter » Thu Jan 03, 2008 4:27 pm

I would like to contribute to the Hardware topic, but I don't think it would be very useful to just list the hardware I use. To provide a useful context for a hardware description, I will also describe the software and procedures I use to make panoramas.

First the hardware: I use a Manfrotto 303SPH pano head on a set of Manfrotto 3021BPRO tripod legs. Manfrotto was formerly known in the United States as Bogen. My camera is a Panasonic Lumix DMC-FZ8. It has a small sensor but a good lens. For a computer I use a locally-built machine with a 1.6 GHz processor and 1.5 GiB of RAM. The tripod and head are overkill for the light camera, but having them means I can upgrade my camera in the future.

I run Autopano Pro under GNU/Linux, with The Gimp as my photo editor. Raw conversion is done by dcraw, sometimes with the ufraw front end.

I have been learning to make panoramas by photographing my garden. I set up the camera and tripod on my front porch, level the tripod and then align the camera in the pano head by observing a nearby object against a distant object in the camera's LCD. If the two objects don't move relative to each other when I pivot the camera, then the camera is well-aligned. Setup, leveling and alignment take only a few minutes.

I set the camera for white balance daylight and ISO 100. The white balance doesn't matter because I will be doing my own raw conversions. The ISO 100 setting is to minimize the thermal noise that is inevitable in a small sensor. I let the camera choose its shutter speed and aperture based on its light meter, and I also allow it to auto-focus. My thinking is that a panorama is not like a portrait, meant to be appreciated with a single glance, but more like a mural, expected to be appreciated by scanning with the eyes and perhaps even turning the head. When you look at a landscape this way, your eyes adjust as you look in different directions, so a panoramic picture should give the same effect.

When I capture the images I use lots of overlap. At full wide angle, when I capture an image every 10 horizontal degrees I get 80% overlap, meaning each item appears in five consecutive images. The pano head holds the camera in portrait orientation, so I can use more vertical separation between rows. For my garden I use a 190-degree horizontal sweep, with three rows at 0, +20 and -20 degrees. or sometimes four rows spaced at 15-degree intervals. The horizontal positions of each row are offset by 5 degrees from the rows above and below, to give the Autopano Pro blender an opportunity to prevent vignetting.

When photographing the garden I must choose a time when the sky is either clear or overcast. Moving clouds causes the lighting to change quickly, and I cannot get enough consistency in the images. Similarly, I must choose a time of no wind, else the trees and bushes will look different in each image.

I capture images in raw+jpeg mode onto the camera's SD memory chip, then copy the data from the chip into my computer, using a directory name based on the subject and date, such as garden_2008_01_03. I divide the files into raw and jpeg subdirectories, then use dcraw to convert all the raw files to TIFF format. The resulting TIFF files go in their own subdirectory.

I run Autopano Pro, and tell it to process all of the files in the TIFF subdirectory, which is much more convenient than picking the approximately 60 TIFF files individually. Because of the large overlap between images, and the complex scene, Autopano Pro has no trouble stitching the images together into a panoramic photograph. When rendering I choose TIFF format, 300 dpi and embedded layers.

If the scene contains large areas which are lighter or darker than usual, the camera will adjust its exposure when recording those images, and the Autopano Pro blender will make a nice transition. However, it might happen that the area is too small for the camera to expose properly while still capturing the rest of the image. If I notice the problem while I am capturing the images I can zoom in on the problem area, in several steps to avoid confusing the Autopano Pro stitcher, and get a well-blended result. Usually, though, I don't see the problem until I have the landscape in front of me. In that case I can use dcraw to create a second set of TIFF files, with exposure adjusted as necessary. Autopano Pro then creates a second panorama, and I blend the two using The Gimp. For small projects I can use the ufraw software to interactively adjust how raw files are converted, but doing this is tedious when you have 84 images in a panorama.

Another benefit of the large overlap I use is the decrease in thermal noise provided by the Autopano Pro blender. I have noticed that I can decrease a picture's noise by taking up to five photographs of the same subject, then averaging each pixel using The Gimp. By using lots of overlap, Autopano Pro does this for me automatically.

I render in TIFF format using embedded layers so I can deal with transients. In a beach scene that includes people and birds, I want a consistent image of a moving object, but Autopano Pro sometimes will create a ghost in place of a moving person. I can choose one of the layers which includes the person's location and either cause the person to fully appear or disappear by making the appropriate pixels opaque.

I approach digital photography from a computer background, so I am sensitive to the need to back up my data. I have written scripts which automate much of the tedium of gathering my picture data and writing it to a series of DVD-R disks. The part I have not been able to automate is removing the written DVD-R from the drive and inserting a blank disk. I have been unable to locate affordable disk changing hardware for which the manufacturer will release the interface specifications.

I would love to read about how others make panoramas, particularly if you use different hardware, software and procedures from mine.
Last edited by John_Sauter on Thu Jan 03, 2008 5:00 pm, edited 1 time in total.

no avatar
hankkarl
Member
 
Posts: 1284
Likes: 0 post
Liked in: 0 post
Joined: Tue Feb 21, 2006 5:32 pm
Location: Connecticut, USA
Info

by hankkarl » Fri Jan 04, 2008 5:26 am

A couple of comments:
1. use a fixed aperture, as small (big f number) as you can without having diffraction soften the image. That way, you get the deepest DoF. On your camera:
the sensor is 1/2.5" (5.76x4.29mm), or a crop of about 6.25 in the horizontal dimension or 5.6 in the vertical.
the lens is a 36-432mm in 35mm equivalent, or 6.5 - 77 mm actual (assuming you use the vertical dimension crop).
the lens has an aperture of f/2.8 - f/8 at the wide end. this gives a DoF of about 3 feet to infinity if you specify a .004mm circle of confusion.
2. use a WB as close as possible to what it actually is. Even in RAW, the less PS work, the better.
3. I use an overlap of 50%, that is, I take about 720 degrees worth of photos for a 360 degree circle. Too much overlap leads to issues with stitching--the edges of the image are more distorted than the center, so two edges that overlap may have some bad points becasue of lens distortion.
4. try using APP with your RAW files. IIRC APP uses DCRAW.
5. With TIFF, use 16 bits for best dynamic range.
6. If you have large EV differences, try shooting manual. shoot one set at 0EV, one at -2EV and one at +2EV. Then use APPs HDR (or Photomatix or similar) and tone map.
7. Exposing to the right will also decrease thermal noise.
8. note that even with RAW, if you brighten the picture, you also increase the thermal noise.

As far as my workflow goes, its similar to yours.
1. Manfrotto 303sph and Canon EOS 5D and Sigma 12-24mm lens. f/16. Fixed WB. ISO 100. There is no detectable thermal noise at ISO 100 on a properly exposed image on this camera.
2. Aperature priority or manual priority, bracketed if I think the pano needs to be HDRed, bracket spacing is 0 -2 +2 EV (I'm leaning toward manual priority now)
when in manual priority, I used fixed aperture and shutter speed.
3. if there are moving objects in image (people, cars, birds) use APP and render three panos, then use PS to remove ghosts, then photomatix to HDR.
3a. If no moving objects, use photomatix on each bracketed set of images, then APP, then render.
Last edited by hankkarl on Fri Jan 04, 2008 5:27 am, edited 1 time in total.

User avatar
John_Sauter
Member
 
Topic author
Posts: 173
Likes: 0 post
Liked in: 0 post
Joined: Sat Jul 28, 2007 4:59 pm
Location: New Hampshire, USA
Info

by John_Sauter » Fri Jan 04, 2008 3:23 pm

Thank you for your feedback. I had tried sending the raw files to Autopano Pro 1.3.0, but it would not handle them. I now see that 1.4.0 does handle raw files from the Panasonic Lumix DMC-FZ8. Handling raw files directly means I don't lose the shutter speed, exposure date and aperture information.

I will switch to aperture priority and F/8. So far I have been taking outdoor pictures, so leaving the white balance at Daylight has been good, but I will change my procedures to use an appropriate white balance based on your suggestion.

I am unable to use 16-bit TIFF because The Gimp is limited to 8 bits per channel, and Adobe Photoshop is not offered for GNU/Linux.

So far I have never had a problem with stitching errors caused by too much overlap. If I do I can remove the bad control points, or remove some images. However, I suggest you try using more overlap than you do. Perhaps Autopano Pro is clever than you think!

I will try the manual exposure bracketing that you suggested.

I don't know what you mean by "exposing to the right". If you mean overexposing, then I agree that the thermal noise should be decreased, but I am concerned about losing highlights.

While the Canon EOS 5D, with its 35.8 x 23.9 mm sensor, is unquestionally a better camera than the Panasonic Lumix DMC-FZ8, you may still have some thermal noise. DP Review was able to see some at ISO 100, though not much. See <http://www.dpreview.com/reviews/canoneos5d/page21.asp>.

Try this experiment: with the camera fixed, take five pictures of a well-lit scene that includes a smooth surface, using the same shutter speed, aperture, etc for each. Pull the images into Photoshop or The Gimp as layers. Align each layer above the bottom layer to the bottom layer by setting its opacity to 50% and inverting it. Adjust the layer's horizontal and vertical position one pixel at a time until the composite is grey.

Set the opacity of the bottom layer to 100%, the next 50%, then 33.3%, 25%, and the top layer to 20%. Zoom in on the smooth surface so that the individual pixels are visible and compare the composite image with any of the individual images. On my camera the decrease in noise is easy to see; perhaps you will be able to see some with yours also.

Again, thank you.

no avatar
hankkarl
Member
 
Posts: 1284
Likes: 0 post
Liked in: 0 post
Joined: Tue Feb 21, 2006 5:32 pm
Location: Connecticut, USA
Info

by hankkarl » Fri Jan 04, 2008 5:04 pm

More overlap = more time, both taking the pictures and processing them.

Outdoors, WB of daylight is good, but if you have cloudy or shadow WB presets, they are good when its overcast. A white balance card is best. But its all bit-twiddling, any of the settings will give you an "ok" picture.

See http://www.llvj.com/tutorials/expose-right.shtml

Thermal noise? you may be able to measure it, but you'll have a tough time seeing it even in a 100% view, and APP does some of the averaging you are talking about. If it is too much to deal with, noise ninja does a great job. And HDR techniques seem to remove a bit of thermal noise also, its not just the averaging, tonemapping often does some smoothing.
Last edited by hankkarl on Fri Jan 04, 2008 5:07 pm, edited 1 time in total.

User avatar
John_Sauter
Member
 
Topic author
Posts: 173
Likes: 0 post
Liked in: 0 post
Joined: Sat Jul 28, 2007 4:59 pm
Location: New Hampshire, USA
Info

by John_Sauter » Sat Jan 05, 2008 4:33 am

Thanks for the tutorial reference; it clarified what you meant by "exposing to the right".

no avatar
Warren Sarle
Member
 
Posts: 53
Likes: 0 post
Liked in: 0 post
Joined: Sat Mar 03, 2007 1:02 am
Info

by Warren Sarle » Sun Jan 06, 2008 12:58 am

hankkarl wrote:Thermal noise? you may be able to measure it, but you'll have a tough time seeing it even in a 100% view,...

What? Klaus was just recently castigating my Nikon D2X for excessive noise levels! :-)

As far as I know, all compact digital cameras have readily visible thermal noise, and the problem is getting worse as manufacturers try to put more and more pixels on tiny sensors.

no avatar
Warren Sarle
Member
 
Posts: 53
Likes: 0 post
Liked in: 0 post
Joined: Sat Mar 03, 2007 1:02 am
Info

by Warren Sarle » Sun Jan 06, 2008 1:16 am

John_Sauter wrote:First the hardware: I use a Manfrotto 303SPH pano head on a set of Manfrotto 3021BPRO tripod legs. Manfrotto was formerly known in the United States as Bogen.

Manfrotto is the manufacturer. Bogen is the US importer.

When I capture the images I use lots of overlap. At full wide angle, when I capture an image every 10 horizontal degrees I get 80% overlap, meaning each item appears in five consecutive images.
...
Another benefit of the large overlap I use is the decrease in thermal noise provided by the Autopano Pro blender.

Which blending method do you use? I can see that linear blending might decrease noise via averaging, but I have never noticed such an effect with Smartblend. I don't know how multiband works and I rarely use it, so I don't know what effect it would have on noise.

When I have stitching problems, it's usually because of too much overlap, but I've never tried that much overlap!

I approach digital photography from a computer background, so I am sensitive to the need to back up my data. I have written scripts which automate much of the tedium of gathering my picture data and writing it to a series of DVD-R disks. The part I have not been able to automate is removing the written DVD-R from the drive and inserting a blank disk.

It's a lot easier to use external hard drives.

User avatar
klausesser
Member
 
Posts: 8836
Likes: 5 posts
Liked in: 64 posts
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany
Info

by klausesser » Sun Jan 06, 2008 4:01 am

Warren Sarle wrote:
hankkarl wrote:Thermal noise? you may be able to measure it, but you'll have a tough time seeing it even in a 100% view,...

What? Klaus was just recently castigating my Nikon D2X for excessive noise levels! :-)

As far as I know, all compact digital cameras have readily visible thermal noise, and the problem is getting worse as manufacturers try to put more and more pixels on tiny sensors.

No, no - i did say THAT :-) !! I said, you stressed the dynamics - and that APP by trying to compensate extreme areas gains levels in dark areas. And THAT causes noise. The differences in your pictures between peaks would have stressed every camera - besides of backs with large pixels and 14/16 bit output.
I often work with a D2x when i shoot people outside the studio - it´s a very good camera and produces very low noise.

Just to correct that! ;-)

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

User avatar
John_Sauter
Member
 
Topic author
Posts: 173
Likes: 0 post
Liked in: 0 post
Joined: Sat Jul 28, 2007 4:59 pm
Location: New Hampshire, USA
Info

by John_Sauter » Sun Jan 06, 2008 8:44 am

Warren Sarle wrote:Manfrotto is the manufacturer. Bogen is the US importer.

The salesman at my local camera store told me that Manfrotto products are no longer re-labeled Bogen when they are imported into the United States. He claimed that this was a money-saving move.

Warren Sarle wrote:Which blending method do you use? I can see that linear blending might decrease noise via averaging, but I have never noticed such an effect with Smartblend. I don't know how multiband works and I rarely use it, so I don't know what effect it would have on noise.

When I have stitching problems, it's usually because of too much overlap, but I've never tried that much overlap!

I have been using Smartblend. Perhaps I should look more closely to see if it is actually removing noise. I learned the value of averaging for noise reduction using The Gimp, and I just assumed that Autopano Pro would do the same.

John_Sauter wrote:I approach digital photography from a computer background, so I am sensitive to the need to back up my data. I have written scripts which automate much of the tedium of gathering my picture data and writing it to a series of DVD-R disks. The part I have not been able to automate is removing the written DVD-R from the drive and inserting a blank disk.

Warren Sarle wrote:It's a lot easier to use external hard drives.

If you overwrite your hard drives, an unnoticed corruption will destroy the backup data. If you don't overwrite your hard drives, backing up gets quite expensive, so you don't do it often enough. A DVD-R disk holds about 2 GiB of data (the raw size is 4.7 GB, but adding ECC reduces that to 2 GiB) and I can get a stack of 100 for about $25.00 US. I haven't done the comparison recently, but the last time I did DVD-Rs were considerably less expensive per bit than hard drives.

User avatar
klausesser
Member
 
Posts: 8836
Likes: 5 posts
Liked in: 64 posts
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany
Info

by klausesser » Sun Jan 06, 2008 2:30 pm

Hey John!

"I learned the value of averaging for noise reduction using The Gimp, and I just assumed that Autopano Pro would do the same."

Noisereduction always is a compromise. So better care for optimal exposures and avoid noise this way.
As an example: if you have - as it was in Warren´s case (and in my eyes caused at least a fraction of the problems) - shots with very bright parts and combine them with shots containing very dark areas by overlapping them, a renderer has to equalize both. Because the pictures were not identical and only differed in lights and shadows but were different angles a real smooth combining is impossible and a compromise is made far from being optimal.
If you sometimes have such situations it´s better to shoot bracketet. Then you can either combine the bracketed shots in Photomatix first and put them into APP then (!) - which also can cause a noise gain - or you select optimal exposures related to the angles you need and use only them to render. Because of them being shot continuously they will match very good using the colorcorrection and maybe using the anchor-settings.

But a reason for some difficulties may be related to Vista´s handling of RAW-formats - i heard of issues with newest Nikon- and CanonRAWs. I don´t know Vista good enough.

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

User avatar
klausesser
Member
 
Posts: 8836
Likes: 5 posts
Liked in: 64 posts
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany
Info

by klausesser » Sun Jan 06, 2008 2:47 pm

John_Sauter wrote:
Warren Sarle wrote:Manfrotto is the manufacturer. Bogen is the US importer.

The salesman at my local camera store told me that Manfrotto products are no longer re-labeled Bogen when they are imported into the United States. He claimed that this was a money-saving move.

Warren Sarle wrote:Which blending method do you use? I can see that linear blending might decrease noise via averaging, but I have never noticed such an effect with Smartblend. I don't know how multiband works and I rarely use it, so I don't know what effect it would have on noise.

When I have stitching problems, it's usually because of too much overlap, but I've never tried that much overlap!

I have been using Smartblend. Perhaps I should look more closely to see if it is actually removing noise. I learned the value of averaging for noise reduction using The Gimp, and I just assumed that Autopano Pro would do the same.

John_Sauter wrote:I approach digital photography from a computer background, so I am sensitive to the need to back up my data. I have written scripts which automate much of the tedium of gathering my picture data and writing it to a series of DVD-R disks. The part I have not been able to automate is removing the written DVD-R from the drive and inserting a blank disk.

Warren Sarle wrote:It's a lot easier to use external hard drives.

If you overwrite your hard drives, an unnoticed corruption will destroy the backup data. If you don't overwrite your hard drives, backing up gets quite expensive, so you don't do it often enough. A DVD-R disk holds about 2 GiB of data (the raw size is 4.7 GB, but adding ECC reduces that to 2 GiB) and I can get a stack of 100 for about $25.00 US. I haven't done the comparison recently, but the last time I did DVD-Rs were considerably less expensive per bit than hard drives.

Hi Warren!

"A DVD-R disk holds about 2 GiB of data (the raw size is 4.7 GB, but adding ECC reduces that to 2 GiB)"

Is that so on PCs? On a mac you have all 4,7GB - i use DVD-DL and they take around 8GB.

"and I can get a stack of 100 for about $25.00 US. I haven't done the comparison recently, but the last time I did DVD-Rs were considerably less expensive per bit than hard drives."

If you work exessively with pictures - as i do (plus video- and audio-editing) - my preference is to have an external sATA backplane and using sATA drives one for each client/Job. The prices are about 85.-€ for a 500GB drive . . or about 40.-€ for 160GB drives. That´s less expensive as DVDs.
Using them hot-swapable it´s very comfortable and fast to handle. The external hot-swapable frame was about 60.-€.

Regarding the time it takes to burn stacks of DVDs and to keep in track where the datas are if you want to have quick access to a decent file it´s a mess to search dozends of DVDs. Even if you correctly label them - you can´t write ALL infos onto the case and so you´ll have to search them one by one . .

Drives with sensitive client-datas are stored in a safe and only put out when used. It´s a matter of second to put the drive into a slot of the frame and to have access to some hundreds of GB.

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

User avatar
John_Sauter
Member
 
Topic author
Posts: 173
Likes: 0 post
Liked in: 0 post
Joined: Sat Jul 28, 2007 4:59 pm
Location: New Hampshire, USA
Info

by John_Sauter » Sun Jan 06, 2008 2:48 pm

klausesser wrote:Noise reduction always is a compromise. So better care for optimal exposures and avoid noise this way.
As an example: if you have - as it was in Warren´s case (and in my eyes caused at least a fraction of the problems) - shots with very bright parts and combine them with shots containing very dark areas by overlapping them, a renderer has to equalize both. Because the pictures were not identical and only differed in lights and shadows but were different angles a real smooth combining is impossible and a compromise is made far from being optimal.

But that is not what I am trying to achieve. I capture each image with exposure set for its own content, rather than finding a good exposure for the important part of the picture and then using it for all images, as I used to do. The result is that the exposure varies from one part of the picture to another, reducing the dynamic range and making a pleasing (to my eyes) panorama.
klausesser wrote:If you sometimes have such situations it´s better to shoot bracketet. Then you can either combine the bracketed shots in Photomatix first and put them into APP then (!) - which also can cause a noise gain - or you select optimal exposures related to the angles you need and use only them to render. Because of them being shot continuously they will match very good using the color correction and maybe using the anchor-settings.

By shooting with automatic exposure, I get the effect of shooting bracketed except the camera chooses the exposure at capture time instead of me choosing it from the bracketed images. However, I will try some bracketed shooting to see the difference. Thank you for your suggestion.
klausesser wrote:But a reason for some difficulties may be related to Vista´s handling of RAW-formats - i heard of issues with newest Nikon- and CanonRAWs. I don´t know Vista good enough.

best, Klaus

Fortunately for me, I am using GNU/Linux rather than Microsoft Windows Vista.

User avatar
klausesser
Member
 
Posts: 8836
Likes: 5 posts
Liked in: 64 posts
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany
Info

by klausesser » Sun Jan 06, 2008 3:11 pm

Hi John!

"By shooting with automatic exposure, I get the effect of shooting bracketed"

No. Definitely not. The best would be you try bracketing and see the difference. Get the Photomatix demo and carefully read the description of what´s HDR good for.

(in short: the point is you extend the dynamic range beyond the capabilities of every camera on earth´s surface :cool: by shooting bracketed and combine the shots in a HDR program. You get a 32bit/ch floating-point HDR-file which you tonemap or tonecompress to have a 16-8bit/ch usual and handable file as tiff, jpg or what else. The advantage is to have the brightest lights and the deepest shadows correct exposed.
There is literally NO situation you can´t handle with that method. For example shooting in a dark cathedral and wanting the painted windows correct exposed and also the dark rows of seats or columns of the sidehall.)

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

User avatar
John_Sauter
Member
 
Topic author
Posts: 173
Likes: 0 post
Liked in: 0 post
Joined: Sat Jul 28, 2007 4:59 pm
Location: New Hampshire, USA
Info

by John_Sauter » Sun Jan 06, 2008 3:59 pm

John_Sauter wrote:If you overwrite your hard drives, an unnoticed corruption will destroy the backup data. If you don't overwrite your hard drives, backing up gets quite expensive, so you don't do it often enough. A DVD-R disk holds about 2 GiB of data (the raw size is 4.7 GB, but adding ECC reduces that to 2 GiB) and I can get a stack of 100 for about $25.00 US. I haven't done the comparison recently, but the last time I did DVD-Rs were considerably less expensive per bit than hard drives.

klausesser wrote:Hi Warren!

"A DVD-R disk holds about 2 GiB of data (the raw size is 4.7 GB, but adding ECC reduces that to 2 GiB)"

Is that so on PCs? On a mac you have all 4,7GB - i use DVD-DL and they take around 8GB.

I have found that DVD-Rs are not very reliable. About one time in 10, reading a DVD-R returns the wrong data with no indication of any reading problems. I do several things to deal with this deficiency:
[ul]
[uli]I write only 2 GiB of data to a disk, and use dvdisaster to fill the remainder of the space with ECC. If I detect an error, and reading again does not correct it, I can use dvdisaster to correct the data using the ECC.[/uli]
[uli]After I have divided my data into 2 GiB chunks, I create a file of checksums, one for the full set of data and one for each chunk. That small checksum file goes onto each DVD-R. When I am reading a 2 GiB chunk of data from a DVD-R I compare its checksum with the one in the file to be sure I read it correctly.[/uli]
[uli]I write two copies of each disk. When dvdisaster finds errors on a disk, it will read those blocks from the second copy to try to create a good image. Combined with the ECC this gives a good chance of reading the data successfully.[/uli][/ul]

klausesser wrote:"and I can get a stack of 100 for about $25.00 US. I haven't done the comparison recently, but the last time I did DVD-Rs were considerably less expensive per bit than hard drives."

If you work exessively with pictures - as i do (plus video- and audio-editing) - my preference is to have an external sATA backplane and using sATA drives one for each client/Job. The prices are about 85.-€ for a 500GB drive . . or about 40.-€ for 160GB drives. That´s less expensive as DVDs.
Using them hot-swapable it´s very comfortable and fast to handle. The external hot-swapable frame was about 60.-€.

Regarding the time it takes to burn stacks of DVDs and to keep in track where the datas are if you want to have quick access to a decent file it´s a mess to search dozens of DVDs. Even if you correctly label them - you can´t write ALL infos onto the case and so you´ll have to search them one by one . .

Drives with sensitive client-datas are stored in a safe and only put out when used. It´s a matter of second to put the drive into a slot of the frame and to have access to some hundreds of GB.

Best, Klaus

Clearly you are working at a higher level than I am. I am strictly an amateur; I derive no income from photography, and so I cannot afford to spend more based on doing more work.

However, let's do some arithmetic. I put 2 GiB of data on a disk, but I write the data on two disks so that's 1 GiB of data for a single-side DVD-R. I get a stack of 100 for $25 US, so that's $25 US for 100 GiB of storage. The exchange rate is currently 1.4786, so 40€ is about $59.14 US. Thus DVD-R gets you about 235 GiB or 254 GB compared to 160 GB if you use 160 GB SATA hard drives.

If you use 500 GB hard drives, 85€ = $125.60 US gets you 5 stacks of 100 DVD-Rs for 500 GiB which is 540 GB. A quick glance at E-bay shows an asking price for a 500 GB SATA II hard drive of $135 US, including shipping. I expect the seller would accept an offer for $125.60.

Thus, I see a significant cost advantage for DVD-R compared to 160 GB hard drives, but little advantage compared to 500 GB hard drives.

DVD-R media is not as reliable as hard drives, but I think my use of redundant storage and ECC makes up for that. I also take your point about hard drives being faster to retrieve data from than DVD-R. That is the price you pay for using a less-expensive backup solution.

Regarding SATA hot-swappable frames, my preference would be to use iSCSI. Ethernet gives me much more cable length than SATA, so I have more flexibility in arranging my space. I hope that someday iSCSI hot-swappable frames will become affordable.
Last edited by John_Sauter on Sun Jan 06, 2008 4:25 pm, edited 1 time in total.

User avatar
John_Sauter
Member
 
Topic author
Posts: 173
Likes: 0 post
Liked in: 0 post
Joined: Sat Jul 28, 2007 4:59 pm
Location: New Hampshire, USA
Info

by John_Sauter » Sun Jan 06, 2008 4:22 pm

klausesser wrote:Hi John!

"By shooting with automatic exposure, I get the effect of shooting bracketed"

No. Definitely not. The best would be you try bracketing and see the difference. Get the Photomatix demo and carefully read the description of what´s HDR good for.

(in short: the point is you extend the dynamic range beyond the capabilities of every camera on earth´s surface :cool: by shooting bracketed and combine the shots in a HDR program. You get a 32bit/ch floating-point HDR-file which you tonemap or tonecompress to have a 16-8bit/ch usual and handable file as tiff, jpg or what else. The advantage is to have the brightest lights and the deepest shadows correct exposed.
There is literally NO situation you can´t handle with that method. For example shooting in a dark cathedral and wanting the painted windows correct exposed and also the dark rows of seats or columns of the sidehall.)

best, Klaus

Unfortunately, Photomatix is offered only for Microsoft Windows and Apple MAC OS. Maybe someday they will awaken to the market for GNU/Linux, as Autopano Pro has.

I see from their web site that it does exposure blending, 32-bit HDR images and tone mapping. Autopano Pro seems to do exposure blending just fine, at least in my limited experience. I don't know what I would do with a 32-bit HDR image. I can't imagine what it would cost to print a photograph with such a large dynamic range. I prefer to reduce the dynamic range of a scene, so I can print it at reasonable cost.

As I look at the Photomatix examples page, I think I see what it is doing. It combines bracketed images into a single image with lots of dynamic range (HDR) and then reduces that dynamic range to 8 or 16 bits to make a printable result. I think I could get the same result by hand by stacking the images, then controlling the transparency of each layer so that each part of the resulting picture is correctly exposed. This is, of course, much more work than using Photomatrix would be, but I believe the results would be the same. Am I missing something?

no avatar
hankkarl
Member
 
Posts: 1284
Likes: 0 post
Liked in: 0 post
Joined: Tue Feb 21, 2006 5:32 pm
Location: Connecticut, USA
Info

by hankkarl » Sun Jan 06, 2008 4:32 pm

Warren Sarle wrote:
hankkarl wrote:Thermal noise? you may be able to measure it, but you'll have a tough time seeing it even in a 100% view,...

What? Klaus was just recently castigating my Nikon D2X for excessive noise levels! :-)

As far as I know, all compact digital cameras have readily visible thermal noise, and the problem is getting worse as manufacturers try to put more and more pixels on tiny sensors.

Sorry, this quote is out of context. I was talking about thermal noise on my 5D. I agree that P&S digicams have a thermal noise problem.

no avatar
hankkarl
Member
 
Posts: 1284
Likes: 0 post
Liked in: 0 post
Joined: Tue Feb 21, 2006 5:32 pm
Location: Connecticut, USA
Info

by hankkarl » Sun Jan 06, 2008 4:40 pm

Warren Sarle wrote:
John_Sauter wrote:First the hardware: I use a Manfrotto 303SPH pano head on a set of Manfrotto 3021BPRO tripod legs. Manfrotto was formerly known in the United States as Bogen.

Manfrotto is the manufacturer. Bogen is the US importer.

When I capture the images I use lots of overlap. At full wide angle, when I capture an image every 10 horizontal degrees I get 80% overlap, meaning each item appears in five consecutive images.
...
Another benefit of the large overlap I use is the decrease in thermal noise provided by the Autopano Pro blender.

Which blending method do you use? I can see that linear blending might decrease noise via averaging, but I have never noticed such an effect with Smartblend. I don't know how multiband works and I rarely use it, so I don't know what effect it would have on noise.

When I have stitching problems, it's usually because of too much overlap, but I've never tried that much overlap!

I approach digital photography from a computer background, so I am sensitive to the need to back up my data. I have written scripts which automate much of the tedium of gathering my picture data and writing it to a series of DVD-R disks. The part I have not been able to automate is removing the written DVD-R from the drive and inserting a blank disk.

It's a lot easier to use external hard drives.

I think the decrease in noise may be as much due to the interpolater as to the blender, but I think both contribute. Its not drastic, or even as much as noise ninja generally does, but it does help.

no avatar
hankkarl
Member
 
Posts: 1284
Likes: 0 post
Liked in: 0 post
Joined: Tue Feb 21, 2006 5:32 pm
Location: Connecticut, USA
Info

by hankkarl » Sun Jan 06, 2008 4:44 pm

John_Sauter wrote:
Warren Sarle wrote:Manfrotto is the manufacturer. Bogen is the US importer.

The salesman at my local camera store told me that Manfrotto products are no longer re-labeled Bogen when they are imported into the United States. He claimed that this was a money-saving move.

A couple of years ago, the salesman told me it was because Bogen's son took over the business, and didn't have the pull to get the products relabeled "Bogen". Nothing has changed but the label. It may be Manfrotto wanted more money to label the gear "Bogen". Its still the same stuff, so if I can save a nickel by having a different name on it, that's ok.

no avatar
hankkarl
Member
 
Posts: 1284
Likes: 0 post
Liked in: 0 post
Joined: Tue Feb 21, 2006 5:32 pm
Location: Connecticut, USA
Info

by hankkarl » Sun Jan 06, 2008 5:04 pm

John_Sauter wrote:However, let's do some arithmetic. I put 2 GiB of data on a disk, but I write the data on two disks so that's 1 GiB of data for a single-side DVD-R. I get a stack of 100 for $25 US, so that's $25 US for 100 GiB of storage. The exchange rate is currently 1.4786, so 40€ is about $59.14 US. Thus DVD-R gets you about 235 GiB or 254 GB compared to 160 GB if you use 160 GB SATA hard drives.

If you use 500 GB hard drives, 85€ = $125.60 US gets you 5 stacks of 100 DVD-Rs for 500 GiB which is 540 GB. A quick glance at E-bay shows an asking price for a 500 GB SATA II hard drive of $135 US, including shipping. I expect the seller would accept an offer for $125.60.

Thus, I see a significant cost advantage for DVD-R compared to 160 GB hard drives, but little advantage compared to 500 GB hard drives.

DVD-R media is not as reliable as hard drives, but I think my use of redundant storage and ECC makes up for that. I also take your point about hard drives being faster to retrieve data from than DVD-R. That is the price you pay for using a less-expensive backup solution.

Regarding SATA hot-swappable frames, my preference would be to use iSCSI. Ethernet gives me much more cable length than SATA, so I have more flexibility in arranging my space. I hope that someday iSCSI hot-swappable frames will become affordable.

The prices are close, but what's your "free" time worth? I'll bet you spend a lot of time being a "disk jockey" with the DVDs. :-)

iSCSI depends on TCP/IP, and is limited by ethernet speeds. eSATA can be 3Gb (vs 1Gb ethernet) and ethernet has a high TCP/IP overhead. eSATA has a 2 meter cable length. SAS (Serial Attached SCSI) give the same data rate as eSATA and has an 8 meter cable length, but comes at a higher price. But the issue may be the drive speed or ability of the PC to deal with that much data.

I'm torn between ethernet attached hard drives and eSATA as storage/backup devices. The ethernet drives are more portable (you can put them on virtually any modern PC or laptop), and all devices on your network may be able to access them, but network traffic will affect the transfer rate. The eSATA are overall faster but require the attached PC to act as a server if you want to access their data from another workstation, or requre that they be moved to an eSATA cage on that other workstation (so your laptop gets less portable).

User avatar
John_Sauter
Member
 
Topic author
Posts: 173
Likes: 0 post
Liked in: 0 post
Joined: Sat Jul 28, 2007 4:59 pm
Location: New Hampshire, USA
Info

by John_Sauter » Sun Jan 06, 2008 6:08 pm

hankkarl wrote:The prices are close, but what's your "free" time worth? I'll bet you spend a lot of time being a "disk jockey" with the DVDs. :-)

That is the big hole in my backup strategy. I have been unable to find an affordable DVD changer for which the manufacturer will release the interface specifications.

hankkarl wrote:iSCSI depends on TCP/IP, and is limited by ethernet speeds. eSATA can be 3Gb (vs 1Gb ethernet) and ethernet has a high TCP/IP overhead. eSATA has a 2 meter cable length. SAS (Serial Attached SCSI) give the same data rate as eSATA and has an 8 meter cable length, but comes at a higher price. But the issue may be the drive speed or ability of the PC to deal with that much data.

I'm torn between ethernet attached hard drives and eSATA as storage/backup devices. The ethernet drives are more portable (you can put them on virtually any modern PC or laptop), and all devices on your network may be able to access them, but network traffic will affect the transfer rate. The eSATA are overall faster but require the attached PC to act as a server if you want to access their data from another workstation, or requre that they be moved to an eSATA cage on that other workstation (so your laptop gets less portable).

At my level, iSCSI drives are plenty fast enough. In fact, at the moment my external drives are on USB 2.0 hi-speed.

no avatar
hankkarl
Member
 
Posts: 1284
Likes: 0 post
Liked in: 0 post
Joined: Tue Feb 21, 2006 5:32 pm
Location: Connecticut, USA
Info

by hankkarl » Sun Jan 06, 2008 6:28 pm

John_Sauter wrote:
klausesser wrote:Hi John!

"By shooting with automatic exposure, I get the effect of shooting bracketed"

No. Definitely not. The best would be you try bracketing and see the difference. Get the Photomatix demo and carefully read the description of what´s HDR good for.

(in short: the point is you extend the dynamic range beyond the capabilities of every camera on earth´s surface :cool: by shooting bracketed and combine the shots in a HDR program. You get a 32bit/ch floating-point HDR-file which you tonemap or tonecompress to have a 16-8bit/ch usual and handable file as tiff, jpg or what else. The advantage is to have the brightest lights and the deepest shadows correct exposed.
There is literally NO situation you can´t handle with that method. For example shooting in a dark cathedral and wanting the painted windows correct exposed and also the dark rows of seats or columns of the sidehall.)

best, Klaus

Unfortunately, Photomatix is offered only for Microsoft Windows and Apple MAC OS. Maybe someday they will awaken to the market for GNU/Linux, as Autopano Pro has.

I see from their web site that it does exposure blending, 32-bit HDR images and tone mapping. Autopano Pro seems to do exposure blending just fine, at least in my limited experience. I don't know what I would do with a 32-bit HDR image. I can't imagine what it would cost to print a photograph with such a large dynamic range. I prefer to reduce the dynamic range of a scene, so I can print it at reasonable cost.

As I look at the Photomatix examples page, I think I see what it is doing. It combines bracketed images into a single image with lots of dynamic range (HDR) and then reduces that dynamic range to 8 or 16 bits to make a printable result. I think I could get the same result by hand by stacking the images, then controlling the transparency of each layer so that each part of the resulting picture is correctly exposed. This is, of course, much more work than using Photomatrix would be, but I believe the results would be the same. Am I missing something?

HDR does combine the images of a bracketed set into one image with a big DR. But just hand-stacking and controling the transparency is a different issue. When HDR is used, some sort of compression or tone mapping must be used for the image to fit into a monitor or printer's color gammut. Tone mapping usually includes local features, so bright areas are darkened and dark areas are brightened, but the whole thing looks good. Many painters (e.g. Vermeer) did this sort of thing.

Attached are three images bracketed by 2 EV, and a tone-mapped result. Notice that it looks "fake" in that a photographer's mind knows that some areas should be darker and some areas should be blown out.








Last edited by hankkarl on Sun Jan 06, 2008 6:37 pm, edited 1 time in total.

User avatar
John_Sauter
Member
 
Topic author
Posts: 173
Likes: 0 post
Liked in: 0 post
Joined: Sat Jul 28, 2007 4:59 pm
Location: New Hampshire, USA
Info

by John_Sauter » Sun Jan 06, 2008 7:34 pm

hankkarl wrote:HDR does combine the images of a bracketed set into one image with a big DR. But just hand-stacking and controling the transparency is a different issue. When HDR is used, some sort of compression or tone mapping must be used for the image to fit into a monitor or printer's color gammut. Tone mapping usually includes local features, so bright areas are darkened and dark areas are brightened, but the whole thing looks good. Many painters (e.g. Vermeer) did this sort of thing.

Attached are three images bracketed by 2 EV, and a tone-mapped result. Notice that it looks "fake" in that a photographer's mind knows that some areas should be darker and some areas should be blown out.

Thank you for the illustrative images. You didn't say what software you used to do the tone mapping, but I think a skilled photographer could get the same result by loading the first three images as layers and then controlling their transparency.

I notice that the tourist visible in the last image appears in the same pose only in the lightest of the input images. I conclude that the area under the arch was taken from the lightest image, whereas other parts were chosen from other images.

The resulting image does not look "fake" to me. It looks like the dynamic range of the scene has been reduced to fit the limits of the screen, but that is not "fake" in my book, because our eyes do the same thing as we look from place to place in a scene. Comparing your fourth image with the first three, it looks better than any of its inputs, which of course is the point.

User avatar
klausesser
Member
 
Posts: 8836
Likes: 5 posts
Liked in: 64 posts
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany
Info

by klausesser » Sun Jan 06, 2008 7:42 pm

Hi John!

"I don't know what I would do with a 32-bit HDR image. I can't imagine what it would cost to print a photograph with such a large dynamic range. I prefer to reduce the dynamic range of a scene, so I can print it at reasonable cost."

A 32bit HDR offers to you an extreme greater tone-range. Especially because it´s floating point - that means, everything is variable and offer you a giant range of combinations beyond standard RGB-tables.
By tonemapping you downsize the 32bit fp to a normal 8 or 16bit int which is used for diaplaying and printing.
The point is the range from which you can build up that "normal" file. You can compare it with downsizing a picture made with a highend-camera and a picture shot with a mobilephone´s camera.
The result of printing even a 5x7" print shot with a highend camera and lens will be definitely better than a 5x7" print from a 1,3MPx mobilephone.

Why? Besides of simply more pixels and a better lens there is also a wider tonerange, your printer can work with. I know, bracketing, DRI, HDR and all around it isn´t easy to understand in deep. But it´s worth being explored!

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

no avatar
hankkarl
Member
 
Posts: 1284
Likes: 0 post
Liked in: 0 post
Joined: Tue Feb 21, 2006 5:32 pm
Location: Connecticut, USA
Info

by hankkarl » Sun Jan 06, 2008 8:39 pm

John_Sauter wrote:Thank you for the illustrative images. You didn't say what software you used to do the tone mapping, but I think a skilled photographer could get the same result by loading the first three images as layers and then controlling their transparency.

I notice that the tourist visible in the last image appears in the same pose only in the lightest of the input images. I conclude that the area under the arch was taken from the lightest image, whereas other parts were chosen from other images.

The resulting image does not look "fake" to me. It looks like the dynamic range of the scene has been reduced to fit the limits of the screen, but that is not "fake" in my book, because our eyes do the same thing as we look from place to place in a scene. Comparing your fourth image with the first three, it looks better than any of its inputs, which of course is the point.

I used Photomatix for this. I also use easyHDR.

I think it would be easier to use the middle image (the correctly exposed one) and tonemap it than to blend three images. But the noise would increase in the dark areas.

If you want to try loading the first three images and controlling transparancy, go ahead.

This brightness illusion is what I mean by "fake" http://web.mit.edu/persci/people/adelson/checkershadow_illusion.html

I specified (or thought I did) that a photographer would think its fake because some of the shadows in the bright part are darker than they should be and the shadowed area is brighter than it should be. Most people would not pick up the "fake" unless they had some experience with color and light. I think the fourth image looks best also, but it does not look like a photograph, assuming one uses a photojournalistic definition. I'd call the fourth image photographic art, but not a photojournalistic image.

BTW, look more closely at the first image--I think the tourist is there also.

edit: Of course, someone like Klaus could probably get an image like that straight out of the camera if he used the right lighting, etc :-)
Last edited by hankkarl on Sun Jan 06, 2008 8:42 pm, edited 1 time in total.

User avatar
klausesser
Member
 
Posts: 8836
Likes: 5 posts
Liked in: 64 posts
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany
Info

by klausesser » Sun Jan 06, 2008 10:17 pm

Hey Hank!

"Of course, someone like Klaus could probably get an image like that straight out of the camera if he used the right lighting, etc :-)"

i´d use bracketing with Photomatix - as i always do when i want a great dynamic range. A simple way is to shoot raw - when there are fast moving objects - and have Photomatix produce 3 bracketed exposures out of them and a HDR-file. That can be then tonemapped or tone-compressed in Photomatix and delivers good results . . . of course not as good as 3 shots of 0/-2/+2 steps.

But it works.

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

Next

Who is online

Users browsing this forum: No registered users and 1 guest

cron