Raw vs. Jpeg

When I'm not walking around with my bell-bottoms and platform shoes wondering why people don't think I'm a cool cat, I'm late to many other things. The debate over whether to capture digital photos in the JPEG format or using the camera's native - called the "raw" - format is one of those. I've mostly ignored this debate and shot JPEG, as it seemed to me that RAW isn't a format that is used for image capture, it is used as a bait to capture nerds in internet photo forum threads. However, I try to challenge my beliefs every now and then, as it is always good to reinforce your own smugness and general feeling of superiority over others.

Which brings me to this.

I started using a DSLR in 2008, which meant using a Nikon D40. That generation of cameras had better high ISO performance than compacts, but the image still fell apart at about ISO 1000 or so. The same went for shadow noise - as one approached the lower 1% of light levels the noise overtook the image. This made shooting RAW a losing proposition. Even if the RAW file held 14 bits of data, that would be seven bits of actual image data and seven bits of random noise.

I shot, and still shoot, JPEG all the time for that reason. Since 8 bits of precision is sufficient to keep up with the sensor, shooting RAW just adds very tangible steps and costs to my workflow without adding tangible benefits. Any 14-bit luminance values that ended up truncated to zero in the JPEG encoding were practically guaranteed to be noise. JPEG is lossy, but I only lost the noise.

However, the newest generation sensors - especially the Sony Exmor[a] series - have very low noise levels and very high dynamic range. This means that the values that the JPEG encoder truncate to zero now contains very usable image information. JPEG is still lossy, but now I'm really losing image data.

The Nikon D3100 is somewhere between the D40 and the current IQ champions D5100 and D7000 in terms of shadow handling in RAW and was what I used to do the comparison. Some may object to me not using state-of-the-art sensor technology for the comparison, but as we'll see, it really doesn't make any difference.

JPEGRAW

This is the full image. It was shot as RAW+JPEG FINE using a Nikon D3100, and I post-processed the JPEG and RAW separately to show the difference in end result, since that is what matters. I've tried to make the two versions look more or less the same, but haven't really spent all that much time doing so. The important part here is the lower right, where the deepest shadows are and where the post processing had to recover the most. Mouse over or click to switch between the result of using RAW and JPEG as the original image. The purpose of this image is to judge the full image from an aesthetic standpoint.

JPEGRAW

This is illustrates the difference in shadow detail. In both cases the crops were taken from the post-processed full image. The purpose here is to show, in detail, what the end result of RAW vs JPEG workflow results in.

JPEGRAW

Here I have brightened the two previous crops by a factor of two. The difference in shadow handling of JPEG and RAW is obvious, with JPEG becoming blotchy while RAW becomes noisy.

What we see here is that the RAW image maintains color almost the whole way down to black, while the JPEG falls apart as we approach the black point. The areas also avoid the blockiness of JPEG. Since the noise hasn't completely overpowered the image the data is more useful. One should however note that the RAW color is interpolated - looking at the colored areas we see that they don't map to any physical features, but that the color information most likely is filled in by software.

What to take away from this, then? Well, RAW has suddenly become useful for some cases - when one needs to capture a high-dynamic range scene and can do so by exposing for highlights, can shoot at base ISO, and can't or won't bracket. All things considered, this is quite a rare event. The traditional situation - capturing a combined indoor and outdoor scene like a room with large windows - is probably better served with exposure bracketing in order to preserve a maximum of interior detail. Exposure bracketing can also extend the dynamic range essentially infinitely as long as you don't mid the time it takes to shoot all those images and fusing them. One should also not ignore the option of using fill flash or other methods of actually controlling the light. But for those situations when you're willing to accept some noise in the shadows and lower mid-range - and when traditional methods of dynamic range control are either unavailable or impossible, RAW is probably the way to go.

For the images above used for testing, this is my conclusion: On one hand I think they show real improvement in the shadows using RAW - but the same improvement could have been achieved simply by waiting for the sun to set some more and it really isn't a big deal. The big deal is that the shadow detail doesn't make or break the images - I don't see one image being clearly artistically superior to the other. In this case my advice is that unless your audience consist exclusively of pixel peepers nobody is going to notice - and in that case you'd be better off blocking all the exits and burning the exhibition building to the ground with everyone still in it.