Lab : Raw, definition, use and workflow

  Before wondering how to develop a raw and to set up a worflow to shoot systematically in raw, let start by defining it and better understanding if you need it … or not ! Sorry, but we will need a bit of thechnique ...

Between sensor and file…
8bits or 12bits, what does it changes ?
Development.
Workflow
Conclusion

Between sensor and file...

Most of digital cameras sensors are built according to a Bayer pattern. Indeed, a single photosite can't see the colors, but only photons. Thus, it is necessary to filter the colors on several photosites, and then to rebuild them by integrating the adjacent photosites to form the full color image. The luminosity, it, is collected by all the photosites (light intensity, or number of collected photons). One sees besides as well as the resolution is not the same one in term of luminosity as in term of colorimetry. There are other systems, like the Foveon sensor, which separates the photosites in various layers perpendicular to the luminous rays, but this technique encounters other problems. Moreover, the eye is much more sensitive to the resolution in term of luminosity than in term of colors.

	On the left, identical resolution in brightness and chrominance. Middle, resolution of chrominance is much lower (exaggeratedly compared to a Bayer pattern), however the resulting image is very close to the first one ! On the right, it is the resolution of brightness which is much lower, catastrophic result.
The eye is much more sensitive to the resolution of brightness than to the resolution of chrominance.

To get a usable image file, it is necessary to demosaice the raw informations from the sensor. That's done through algorithms, different according to whether one uses the DSP (Digital Signal Processor, like Canon's Digic) inside the body or an external derawtiser, each algorithm having their pros and cons (sharpness, colors, moire, noise, etc.).

Inside DSLR body, electronics can generate a raw file directly, or send the data to the DSP which will demosaice sensors datas and apply in-body parameters in term of white balance, saturation, contrast, etc, and will thus provide a jpg file already treated. On other hand, the raw file, to which no treatment was applied, will have to be demosaiced on a computer with a derawtiser software, which will make it possible to fine adjust each parameter, contrary to in-body adjustments for jpgs.

 Top

8bits or 12bits, what does it changes ?

Extremely relevant question, since the enormous majority of the screens and printers is unable to restore more than 8bits, in a rather restricted gamut (color space). It is however there that goes one of the most important advantages of the raw, which will allow a treatment on white balance, saturation, contrast, exposure and such much more thorough than in jpg. To know that digital backs captures directly in 16bits, last Canon 1DIII/1DsIII and 40D in 14bits, and other DSLR in 12bits.

A pixel of an image is composed of three layers of colors, Red Blue Green, each layer being able to contain a certain number of tones. In 8bits, each layer is thus coded on 8bits (or 8bpc, for Bits Per Channel), that is to say 256 tones of red, green, and blue. Indeed, data processing being binary, one has 8 “0” or “1” to code information, so 2⁸ = 256 possibilities. On three layers, one thus obtains 256³ = 16 777 216 colors, that is to say 24bpp (Bit Per Pixel).  For example, in 16bits, there is 2¹⁶= 65536 tones of colors by layers, that is to say 655363³ = 281 trillions of colors, or 48bpp. For information, the human eye distinguishes around 12 million colors, our image 8bpc is thus enough, as final format.

The fact of working on 12bits or more in raw allows, in term of exposure but also of tones, to choose in these 12bits, the 8bits we will keep, or even compress these 12bits in the highlights and shadows to make them “fit” in the 8bits of our final format. It is what makes possible to recover blown highlights and/or dark shadows, while working on the curves. Because even if we do not "see" these informations on the screen, they exist in the file, and the derawtiser precisely allow to shift these infos in the visible ones, offering us the choice ! Whereas on a 8bits jpg, this work has already been done with more or less of success by the DSP, you thus do not lay out any more additional information apart from these 8bits. The jpg thus contains only part of the information collected by the sensor, contrary to the raw.

Jpg compression

Even with maximum quality, the jpg is always compressed, and by destructive algorithms (contrary to the zip/lzw of the tiffs). There is thus here still, loss of data, which will make more delicate any treatment on the file. It is nevertheless not the most important aspect, as with minimal compression settings (check your inbody settings), the losses are tiny.

 Top

Development.

The RAW is thus a rough file from the sensor, not demosaiced (Bayer pattern), not interpreted by the in-body DSP. It is thus free from any treatment of sharpness, saturation, denoise etc. It is coded on 12bits (14 or 16 on certain bodies), contrary to jpg files where the values are compressed into 8bits by the DSP according to a curve specific to the manufacturer. The jpg is thus only a part of the raw, left chosen by the DSP, not by you ! The raw offers true digital negative, that can be developed with a derawtiser by adjusting exposure, white balance, tonal curves, saturation of certain colors, contrast, etc, with its own please !

So, you will have to carry out the tasks carried by the DSP when working in jpg. For that, many tools exist, named derawtisers or converters, or simply raw softwares.

Raw converters

It thus will be necessary to find a software, to demosaice (develop) your raw files. If the choice were rather restricted a few years ago, it is plethoric today, each one having its advantages and disadvantages, but in the end it is rather a question of taste and/or cost !
Here is a list, nonexhaustive, of well known raw converters. Keep in mind that most of DSLR manufacturers provide their own converter, like DPP (Digital Photo Professional) for Canon or CaptureNX for Nikon.

	Software	Editor
	Digital Photo Professional	Canon
	CaptureNX	Nikon
	LightRoom	Adobe
	Camera Raw (ACR) (inside Photoshop)	Adobe
	Capture One C1Pro	Phase One
	Apple Aperture (Mac only)	Apple
	Breeze Browser	Breeze Systems
	DxO Optics Pro	DXO
	Bibble	BibbleLabs
	RawShooter Premium/Essential	Pixmantec (bought by Adobe, see Lightroom)
	Silkypix	Silkypix

I personally tested all the converters above, simply to be sure not to be mistaken according to my use. After having used CaptureOne which is excellent in term of demosaicing but a little just in functionalities, I used RawShooter, which offered more functions and was very fast and intuitive. I since switched to Lightroom, which is simply excellent in term of possibilities, a true darkroom, but which also manages your files with functions of cataloguing. It is also the one that enables me to go much further under development and almost not to need photoshop anymore. For the moment, I do not see what will make me change ... But, it is my opinion, completely personal, and I can only advise you to make your owns !

Non destructive modifications

Another large difference between raw and jpg, is that in a converter, all adjustments of development are nondestructive, i.e. they do not affect the file, as long as you don't export the file in another format. Indeed, the converters don't save nor modify a file in raw format. What wants to say besides, that the raw has a certain form of authenticity, as far as you can't save a file in this format. But that wants especially to say that all adjustments are in fact an interpretation of the raw file. For example, Lightroom records for each raw a file named sidecar, in .xmp, which contains all these adjustments (except if you decide to convert them into dng, which contains the sidecar inside). Of course, in jpg, you can safeguard your adjustments as multi-layers psd, but the weight of the file increases by as much, and especially the file changes between each stage. Concretely, that wants to say that if for example, you modify the white balance on a jpg in the blue ones, then again modify it in the reds, the second modification is done starting from the “blue” file, already faded ! Whereas in raw, it is simply an adjustment, that you can modify tons of time without never fading the data of the file ! It is a really important point during the development.