In the heady, fume-filled days before the advent of digital cameras, a major polarising choice for the photographer was between negatives and transparencies. They require a slightly different technique and approach, so tend to be used for different things. Transparencies are particularly good when you need accurate colour, but they are very unforgiving of exposure error and lack some of the flexibility of a negative / positive process.
These days, assuming for the moment the you have made the decision to shoot digitally, the roughly equivalent choice is between JPG or RAW capture. So what is the difference? Is one fundamentally better than the other, and if not, how should you choose between them?
To answer these questions it helps to have a basic understanding of how digital cameras capture and store image data.
At he heart of almost all current digital cameras is a light-sensitive chip (the sensor) that takes the place of the film in a conventional camera. The surface of the sensor consists of an array of light-sensitive elements or pixels (known as photosites) arranged in a rectangular grid. The number of photosites that the chip has is the camera’s resolution, usually expressed in megapixels. Fuji sensors use a diagonal array of hexagonal pixels, but for our purposes this difference is minor.
The sensor itself is a monochrome device - each photosite records only differences in brightness, so it is roughly equivalent to B&W film. To record a colour image a camera needs to record a red, a green and a blue image of the scene. A film camera does this by using film with with 3 layers with red, green and blue filters between them (a scheme that is replicated by the 3-layer Foveon sensors used in Sigma cameras).
[singlepic=36,150,150]
Most digital cameras (Sigma excepted) encode colour information into the monochrome image using a Bayer Mask Filter - a fixed pattern of red, green and blue filters placed over the sensor’s photosites. This causes each pixel to record either a red, a green or a blue value. To produce a full-colour image the missing colour information must be filled in for each pixel.This is done by estimating the missing colour values for each pixel from the values of the surrounding pixels of the missing colour - a process known as demosaicing. A red pixel will be given blue and green values based on the levels of blue and green in the surrounding pixels. Thus the colour (chrominance) resolution of the image is slightly lower than the brightness (luminance) resolution. For most purposes this difference is not perceptible - indeed, image and video compression schemes often exploit this fact by compressing colour information more than the brightness information (chroma sub-sampling).
When you press the shutter on a digital camera, this Bayer Masked data is captured by the chip and passed to the camera’s processor. Here, one of 2 things can happen: Either the data can be written to the camera’s memory card unaltered (this is a RAW file) or it can be processed into a JPG file and then written to the memory card. In some cameras there is an additional 3rd option where both files are written.
You might think that it would be much faster to just write the raw sensor data straight to the memory card rather than doing all the processing required to make a JPG first - however, this is not the case. The internal memory and processor in a digital camera are much faster at compressing data than current flash memory technology is at storing it, so the camera can make a jpg and write it to a flash memory card in much less time than it would take to write the raw data. This internal compression also frees up space in the camera’s buffer, allowing more shots to be taken in a burst.
It should be clear, therefore, that there considerable speed advantages to be gained from shooting JPGs. So why not just shoot that way all the time?
To understand that we need to examine the JPG file format a little. JPG stands for Joint Photographic Experts Group, and is an industry standard method for storing compressed image data - that is to say, storing images in smaller files that take less space and can be uploaded and downloaded more quickly. JPGs are very good at this, but to achieve such high compression ratios the JPG algorithm has to be what is known as “lossy”. This means that data is not just stored more efficiently (as in “lossless” formats like TIFF), with JPG compression some of the data is actually thrown away. The file you get back when you view a JPG is not the same as the file you compressed. It might look the same, but it isn’t. The reason you don’t see a difference is that the JPG algorithm throws away detail that you probably wouldn’t see anyway.
This is both the strength and the weakness of the JPG format. The JPG engine in the camera has optimised the image for the way it is currently displayed. In the process it has discarded shadow detail, dynamic range and colour information. It has created a lot of problems in the image, but has hidden them neatly under the carpet. If, at a later stage, you want to alter the image in some way, you may end up exposing some of these hidden problems. The most obvious example is brightening an underexposed image and finding noise or compression artefacts hidden in the shadows. A less obvious impact is in subsequent B&W conversion. The eye is not equally sensitive to all colours, being more sensitive to green and relatively insensitive to blue. The jpg algorithm exploits this peculiarity by compressing the blue channel more than the others. If you use a lot of the blue channel data in a B&W image conversion it may be noisier than you expect.
So which should you use? It depends entirely on what you are doing.
JPG is a great output format. A well compressed JPG is practically indistinguishable from the original image, and is a lot smaller. When uploading images to a lab for printing I have no problem with using a low compression JPG. If you’re shooting images that will go straight to print or will have minimal post-processing, this is almost certainly the way to go.
RAW files on the other hand, contain a lot of extra information, giving you more headroom for adjustments. They also preserve the original image data, before the demosaicing process is run. As this processing adds estimated colour information to the file there is more than one way to do it. Different RAW file processing software, therefore, tends to produce slightly different results. Effectively, you can experiment with different developers on a digital file. So when the additional size and write times are not an issue, and there will be sufficient time for the required post-processing, RAW files are hard to beat.
The pragmatic approach would suggest that if you’re using JPGs and don’t see a problem, you should stick with it. But it’s worth experimenting with both, as it’s not uncommon for a computer-based RAW conversion to be slightly better than a camera JPG, as the computer has more processor power to apply to the job and doesn’t have higher camera frame-rates as a design goal.
