Full 16 bit workflow
Creating Panorama in full 16 bit workflow
One of the biggest problems in panorama creation is dynamic range. In order to save both maximum number of levels and maximum dynamic range it is recommended to use a 16 bit workflow. 16 bit refers here to the single channel bit depth. However, some applications and tutorials use the term '48 bit' which refers to the complete bit depth for all channels and is essentially the same.
This tutorial is for the advanced user only. It assumes you are familiar with shooting, stitching, blending 8 bit images and with basic and advanced editing techniques in photoshop like using image layers, adjustment layers and masks.
To use 16 bit workflow you need a camera capable of saving raw files or a scanner plus application that allows to save 16 bit.
A recent panotools (pano12) version supports 16 bit as well as the usual GUIs. If you want to use autopano you might need ImageMagick in order to batch convert images to 8 bit since autopano currently has a bug that causes an error message for some types of 16 bit TIFFs. If you want to edit the layered stitching result you will need Adobe Photoshop CS This is not needed if you use enblend for blending the images.
Photographing the panorama
Using a digital camera
To get maximum dynamic range with a digital camera expose for the brightest highlight you want to have details in. In a sunlit landscape this are usually the clouds but could as well be a white building. Digital cameras clip the highlights but not the shadows.
Shooting in RAW allows to generate 16 bit tiffs from these files, with precisely adjusted exposure and white balance, or even different exposure and white-balance settings to assamble the images again in a more favoring way, eg another white-balance for the windows of interior-panos.
Converting Raw-files with the Adobe raw-plugin offers the most adjustments to my experience. Chromatic aberrations and vignetting can be adjusted, and its possible to simulate exposure-bracketing from a single exposure, to use tricks like contrast blending or photomatix.
Speeding up the workflow can be easily done with dr.browns image-processor. A detailed tutorial-movie and the image-processor is available here: http://www.russellbrown.com/tips_tech.html
Using analog film
With analog color negative film expose for the darkest shadow you want to have details in. A tripod is recommended, since the exposure times might get very long especially if stopped down. I use a typical exposure of 1/15s at f/16 in bright sunlight with Fuji Reala. Color negative film clips the shadows but not the highlights. It is a good idea to shoot an extra image containing some color chart or at least a grey card plus a white piece of paper and something entirely black. This will help to white balance later.
Avoid any clipping while scanning. Switch off any automatics in your scanning software that influences white balance or exposure. Save as 16 bit TIFF (somtimes refered to as 48 bit TIFF). You got to xperiment with your scanning software.
With Vuescan use advanced workflow (see manual). Set color balance to 'None' on color tab. Don't worry about the results beeing dull and having a color cast. Save as '48 bit TIFF'
If you want to use autopano it is a good idea to work with 8 bit copies of your images for stitching. Autopano should work with 16 bit images but currently refuses to work with certain types. Stich your image as ever. Once you finished with autopano you can revert to the 16 bit originals again.
Older versions of PTGui silently degraded images to 8 bit if you choose a cropping frame outside the image bounds or if you used a filter.
Basically there are three possible ways to blend you images:
- let panotools do anything by choosing any single image output format like TIFF JPG or PSD.
- Output multiple TIFF files and blend manually in an image editor using layers. Currently the only program I know that handles 16 bit layers is Adobe Photoshop CS
- Using enblend
Enblend handles 16 bit images well. Do what you ever did using enblend and you finish up with a blended 16 bit TIFF. However, be aware of the additional alpha channel enblend puts into the image.
Choose multi image TIFF as output, either with feather (TIFF_mask) or without (TIFF_m). Use my 'Import TIFF files with alpha channel to masked layers' action you can download from http://www.erik-krause.de/ttt to assemble the single images (usually named <yourimage>0000.tif, <yourimage>0001.tif a.s.o) into a layerd Photoshop document.
If you have shot a geycard previously, load this image as a separate file and place a levels adjustment layer above. Use the grey eyedropper to get neutral mid tones but don't use the white or black eyedropper as they might cause clipping.
Open the Info palette. Set a color sample point (5x5 average) to the black and white image detail and adjust input white and black point for the color channels manually in order to get equal color values for all channels. Check the RGB values for the black, the white and the grey sample frequently.
Try to get the minimum common color value for the white point (by not changing the channel with the highest value) and the maximum common color value for the black point (by not changing the channel with the lowest value). Most probably you will have to repeat these steps until you get neutral black, white and grey since any adjustment influences the others.
Once your'e done drag the adjustment layer from the layers palette to your pano on top off all other layers. If you used a grey wedge you can make an even finer adjustments using curves instead of levels and adjusting several shades of grey to be neutral.
If you didn't shoot a greycard, simply place a levels or curves adjustment layer on top of the image layers and adjust colors to look naturally.
Use additional adjustment layers to adjust contrast and / or brightness. It might be a good idea to use a s-shaped curve in order to increase mid tone contrast. Save the document prior to flatten to be able to correct without loose of levels.
After that edit the individual layer masks to avoid visible seams and eventually use additional adjustment layers grouped with single image layers to adjust color or brightness differences.
If you want to make the image look more brilliant, flatten to background and use unsharp mask with threshold 0, amount from 20% to 50% and a radius between 70 and 250 pixels (depending on image size). It's a good idea to reduce to intended output size prior to this step since it is very slow on large images.
-- Erik Krause 05:56, 18 Mar 2005 (EST)