After choice of type of equipment and major hardware/software vendors, there are still two choices to be made to complete one’s PhotoID: camera output file format and image color space.
File format: RAW vs. JPEG or Both (or even TIFF)
RAW is the only image format that enables display of all that the camera saw. RAW format is often considered the negative for the digital age. Each vendor of prosumer and professional cameras offers a proprietary RAW format, using the TIFF/EP standard (ISO 12234-2), itself based on a subset of the proprietary TIFF standard together with a subset of the JEITA Exif standard. There is a variety of software that can render RAW images from various vendors directly into displays and prints, without invoking an intermediate file format such as JPEG.
Although some cameras offer an ‘either’ or ‘both’ choice between JPEG and RAW, I have not been able to conceive a workflow that is improved by having both. So the following addresses the ‘either’ choice. JPEG may be the final choice for many, for instance those with sufficient photographic skill to use images direct from the camera, and those with no pressing need to squeeze the last ounce of color performance from one’s photos. For the rest of us, in particular those who usually edit our images and don’t want to limit the image color performance in any way, RAW is the choice to best support our style. TIFF is also offered on my camera, but it is only 8-bit, and would seem of value only to a photographer with a required workflow and tool set customized around processing TIFFs.
I see three considerations in the choice between JPEG (suffix .jpg) and RAW (Nikon suffix: .nef):
1. Image Size (not pixel dimensions, but the information stored for each pixel)
The finest quality JPEG image is only ~25% the size of the full quality corresponding RAW image. Lesser quality (higher compression) JPEG files can be 100 times smaller than the corresponding RAW file. A compression algorithm squeezes out all those bits of information that the algorithm’s designers claim are not visually important. And these designers are close to right; one would find it difficult to distinguish a RAW from a fine JPEG image reproduction under casual inspection. A simplification in workflow also usually results from choosing JPEG as one’s camera output format, but there are some qualifications to consider below.
If file size is a concern, but RAW seems to best support one’s workflow, smaller RAW file sizes are also options, which keep all the benefits of RAW processing below, but with smaller pixel dimensions. If one only makes 4″x6″ prints of images, such a file size benefit might be worthwhile. But when aiming for large prints, maximum resolution is called for.
2. Quality Preservation under Editing
RAW format offers non-destructive editing; the file is saved along with an edit list. Each time the file is opened, the original image is re-rendered according to the embedded edit list. Even if only occasional touch-up is done, capturing the image in RAW format may be desirable. And since I edit virtually every photo, RAW is a no-brainer.
JPEG editing is destructive, changing the original file. Thus to keep an original copy, the original must be saved somewhere, requiring two copies floating around. And if one ever reverts to the original, any needed editing would need to be re-done. Finally, every time a JPEG image is saved, its quality drops a little.
JPEG editing can be made non-destructive if one first converts the JPEG image into a RAW format and then uses a RAW format editor. Thus, I use Nikon CaptureNX2 to save the JPEG image in Nikon RAW format prior to editing.
Any images to be posted to the web, or to be exported in portable slide shows, will need to be rendered into JPEG format at some point in one’s workflow. In such cases, the lost quality will not likely be missed.
One caveat with RAW format is restriction to editing tools that understand the specific RAW format; almost all do. A related caveat is that once a RAW file is edited, the edits are recognizable only by the program suite that did the edits. Thus, for example, neither iPhoto nor Aperture will ‘see’ the edits in a RAW (NEF) file edited by CaptureNX. These programs see only the RAW negative and its embedded preview representations. They then allow you to edit this negative within the application.
Once you settle on a software suite for RAW processing workflow, then editing, display, and printing can all be performed in the RAW realm.
3. Color Depth
JPEG limits color depth (sample size for each pixel color represented). My camera and similar use 16-bit color depth per channel for internal processing, and the RAW format captures nearly all of it. JPEG only supports 8-bits per channel. Thus JPEG throws away a lot of color information related to the number of gradients (shades) available within a chosen palette.
Whether this matters depends on the exposure characteristics of the image and the output type; in many cases it does matter, particularly on high ISO images, or underexposed areas of an image. Such images may require you to boost the exposure in post editing, and extra color depth may provide vital detail information. While no definitive testing can be quoted, there are hints that the signal-noise ratio is improved in these dark areas using full color depth.
Increasing color depth also may ensure the sample size is sufficient for the color gamut selected. If a very wide gamut is invoked for image rendering, limiting oneself to 8-bit color sampling (256 gradients) may produce posterization in the image, due to the color range represented by each sample being too large (see below).
For the majority of generally non-problematic exposures, the question of appropriate color depth may more revolve around whether the output device can see more than 8-bit depth. For on-screen display, most all current in-plane switching (IPS) LCD displays respond to 8-bit color, but the video card may be able to drive the display via dithering of extra color depth bits. Displays in the future will support higher color depths. So more than 8 bits per color can already make a subtle display difference, and perhaps more so in the near future.
For color prints, there are high-end photo printers with 16-bit capable drivers (Epson professional models from 3880 up), and Snow Leopard supports them. For the best color experience on monitor or printer, choose RAW image capture.
Nikon offers an additional RAW format choice: 12-bit, or full response 14-bit color. The full color response increases file size by 30% and slows down processing on some camera models. I doubt if the difference is noticeable in any workflow I will employ, but my fast computer and large disks permit me the luxury of going for broke in color depth. And for that rare, deeply underexposed image area, those extra 2 bits may mean the difference between detail and dark void.
Color space: Wide-gamut vs. sRGB
My camera offers a choice of color space: sRGB and the wider gamut Adobe RGB (1998). sRGB is a Microsoft standard that originally matched the set-up of the majority of CRT displays, and also became the de facto Internet standard. But the ‘color management’ aware user will get more color information out of Adobe RGB in today’s world. The only reason to remain in sRGB is for easy backward compatibility with the old world. sRGB uses ~35% of a reference color palette. Adobe RGB uses ~50% of this palette.
The wider the color gamut, the wider the range of color that must be represented, with the caveat that using a wider gamut may require increasing color depth to avoid the posterization, due to a wider gradient being represented by each color value. Adobe RGB can just get by with 8-bit color, but post-processing can benefit from greater depth (the difference between hundreds and millions of gradients per sample is significant). Since I do substantial post-processing and want to have the most colors to work with, I choose Adobe RGB, whose gamut matches the default calibration of modern Mac LCD monitors: gamma = 2.2; white point = D65.
Color space selection is imposed in camera if the selected output file format is JPEG or TIFF. In these cases, since the camera is producing a fully rendered image, the full range of colors (the sensor color space) is clipped to match the selected output color space.
Alternatively, if a RAW file is specified, no color clipping occurs to the sensor-derived values; they continue to represent the full sensor color space. In the RAW case, instead, the color space selection is recorded in the file’s metadata as a processing intent. Subsequent file rendering will perform the specified color space clipping per recorded intent.
(Aside: A RAW file does contain two embedded JPEG image copies, a small thumbnail and a full size preview file using basic level of compression. Both of these images do have their color values clipped to the selected color space, but neither of these embedded images has an ICC profile embedded within to identify the color space in which they were encoded.)
Fully modern browsers are color managed. A JPEG file with Adobe RGB ICC profile embedded will be rendered correctly in such browsers, to within the capability of the operating system and one’s display monitor. Safari and IE9 are compatible with ICC version 4 color profiles. Firefox is compatible with ICC version 4 color profiles, but it is not the default. One needs to edit the about:config parameters for gfx.color to turn on version 4.
Microsoft was late to this game. Earlier IE versions were not color managed; they simply assumed the whole world is sRGB, so an Adobe RGB image may appear with compressed colors. If you will use a workflow based on a more expansive color space than sRGB, choose a browser that can handle your color profile. Google is even later to the game. As of 2012, Chrome is not a color-managed browser.
Note that these comments refer specifically to Internet display of photographs for their own sake, which are probably being rendered on a neutral color background on a web site. Other times, web site developers may choose to embed color images as part of a site’s basic visual format. Since HTML- and CSS-encoded colors reference values in an sRGB palette, images selected for such use should best be rendered in sRGB to benefit from intended color matching.
When choosing a color space other than sRGB, it is necessary to be consistent in embedding the correct ICC profile within the image file at each step of the workflow. Once one establishes a correct workflow, then use of other color spaces becomes as glitch-free as using the more compliant sRGB space everywhere.