Do we always need Big Honking' Images?

While having a 18 megapixel sensor camera like the Canon 7D and being able to produce 21 megabyte images brings much to the table in terms of print size, crops and creativity. Not all images require such firepower.

It's not a question of always having what's on the top shelf, but more so having the right tool for the job. At full resolution, the 7D is capable of printing a decent 24×36 image. However, how many of your images require that level of resolution? When it's a personal choice and you just so happen to have access to a 44" wide printer, then why not? However, in my busy day, not every press of the shutter requires the end result to be a 21MB image, when my intended deliverable is something more like an 8×10 placement in a brochure, or a post to the web.

Trimming 44" posters, shot on a Canon 40D in full RAW, for a sales event.

Proper image memory management by "Intent".

It makes sense to regulate the size of your images based on their final intended use, and while it seems silly to worry about this when hard drives and memory storage is vast and cheap these days. That computers and software are also much faster, and if you just shoot jpegs, where the files are incredibly small compared to a RAW. It can still make a difference for the busy professional with regards to time and the sheer volumn of the images he or she has to deal with.

So, while I agree that hard drives are much cheaper. We still have to factor in maintenance and the archiving system that remembers which drive holds a particular image when it's needed. Furthermore, full RAW images today are much larger than a few years ago and continue to bog down even my Dual Quad Work station.

I also have friends whom shoot full resolution for everything, and have drawers full of hard drives to prove it, and the headaches of finding their images and having to replace those drives when their warranty expires (including the need for duplicate drives as a backup. It can get pricey). In comparison to my having only ONE drive that houses some 19,000 keepers, all in RAW format, and simply moving to a larger drive as needed (and I do have and recommend keeping a duplicate drive for backups as well).

Of note, I come from a film background and I have the a habit of taking fewer shots, and when shooting 6 frames per second, only keeping the one shot that matters (rather than all of them).

RAW verses Jpeg.

So why even bother with RAW to begin with? RAW files compared to jpegs are huge.  I know that a 3600 pixel, 3.5mb jpeg at a quality of 76, and 240dpi will make an 11×14 print that easily matches one generated straight from an uncompressed 21mb RAW at 300 dpi.  It's the value of diminishing returns, or overkill.  However, I need the advantage of the extra color space that RAW's provide for editing, and editing 8bit jpegs is a destructive process as they break long before you hit the limits in a RAW file.

While RAW is an advantage, all those full sized 21+MB images can be a challenge. Many of us out in the field are using Lightroom on laptops with media on external USB hard-drives.  It's a lot of memory to be throwing around, and capture and edit sessions can slow significantly as a result.  Especially when you have impatient clients looking over your shoulder.

It's seems feasible that we do some research into sRAW as an alternative. After all, I managed well in the last five years with both a 11MP Canon 1Ds, and a 10MP Canon 40d for all but my largest prints.

Fortunately, I am not the only one whose confronted this issue, and in response, Canon has provided a solution in form of small RAW files, "small RAW" (sRAW) at 4.5MP, and "medium RAW" (mRAW) at 10MP. Where I can selectively choose the image resolution for the any given shoot and still enjoy all the benefits of editing RAW files.

Files sizes of the 4.5MP sRAW average about 8.6MB, comparable to what I had with the 40D, and the mRAW at 13.5MB, and full RAW at 21MB.  Obviously, working with sRAW files in a project were just as quick and responsive to what I had experienced with the 40D, just as quick as working with jpegs.

However, before I simply incorporate sRAW into my workflow, other than image size, what exactly am I losing here?

If you want a better understanding of the math, read it here (pdf) . Knock yourself out. I'm more of a hands-on guy, and the eyes do a fairly good job of estimating the result.  That's what matters.

Shooting Color Charts, seeing the real differences.

Typical MacBeth color chart.

Other than lower resolution, what about the rendering accuracy of sRAW vs full RAW?  That is, does less data mean we have less color, and if so, how much?  In order to find out, we have to shoot what is known as a Macbeth Color Chart. Shooting the chart at each of the three RAW sizes in the exactly same setting, and then compare the results to check for inaccuracies.

The Macbeth color chart is a sampling of basic color values used for calibration of camera's and systems (useful in multi camera shoots).  While the correct use of the color chart can seem confusing, it really isn't.  It's simply a known reference to use as a baseline.  In our test here we're not after color profiles to correct color, but to determine the value difference in those colors between sRAW and full RAW, and hence determine if the loss is significant enough to reject the usefulness of sRAW.

A typical two light setup for an evenly lit color chart

To do this, we carefully shoot a chart in a known, uniform lighting condition, in this case, 5000k studio flash.  We could also shoot this outside on a cloudy day, but would have to ensure even lighting across the entire piece and introduce unwanted variance.

I took three shots each, with two Canon 7Ds, at sRAW, mRAW and full RAW. All other settings remain unchanged.

Above color correcting the charts in Lightroom.

In lightroom, using the full RAW capture, 1) I picked the neutral gray patch for white balance, 2) then I set the black point to zero, and 3) increased the white patch luminance to 100%.  I then copied these values to the other charts and then compared my findings.

Interestingly enough, the difference was so minute, that it could only be seen in software at the pixel level.

Above, left to right:  sRAW, mRAW, and Full RAW (click to see larger size).

To better understand the difference, I then loaded both the sRAW and full RAW files as layers in Photoshop, subtracting one from the other, and provided a gray scale image of the result.

Above: sRAW vs full RAW color differences, (grey areas reflect differences).

Looking at the chart above, the shaded area's represent the differences in color shift.  A perfect match would have been completely white image. 

Upon careful inspection at the pixel level, most of the differences (gray areas) attributed to noise mismatch, something unavoidable with digital sensors.  However, it took this extreme to see the real difference, and that difference is so faint that its hardly a failing in using sRAW.

I also noted some edge ghosting of the sRAW image.  Which makes sense, as the full RAW image allows for more information to describe any abrupt transitions in color.  Likewise, you will note that the black patch does not have any edges at all.  But the edges of the card and the stand do.

Above: Comparing the visual differences between sRAW and full RAW (click the image to see a larger version).

While I'm fine with how sRAW handles colors, I'm a bit skeptical about the edge issues.  I did not sharpen the charts by any means, so sharpening would certainly bring out that edge and using sRAW may become an issue with shooting for keys or green screen for masking. Use full RAW if your capture is with green screen.

Regarding color correction. I remember years ago spending hours learning how to make a custom profile in Photoshop for my 1Ds mark I before Adobe released their custom profile tool.  While we're not going to cover that here, I've provide the link in case your interested.

The proof is in prints.

Viewing images on-line at can be quite forgiving, and only by printing, which requires three times the resolution, can one see where the image fails, and thus determine the usuable limits of sRAW.

Therefore my next test was to set up a scene filled with various devices that represented the bulk of my work. Medical implements with shiny metal surfaces, transparency, and fine text with sharp tiny objects. All shot against a white sweep. While certainly not a beauty shot, it was something I quickly threw together over lunch the with intent on checking perceived sharpness, dynamic range, burn outs from reflections, and soft text, and pixelized gradients.

I then shot the scene on a tripod, tethered to my computer for remote shooting to remove any user error from the result. I also allowed shadows, rather than a complete blow out, to see how the gradients were handled.

Next I printed a sample of each RAW resolution, assuming an 11×14 crop, and a second scaled up 200%, and then noted the differences.


The real deal, comparing resolution on prints

When viewed at arms length, all three resolutions held up admirably for 11×14.  The sRAW did well and only with a loupe could I see any minute difference.  However,  at 200% the sRAW fell apart, with the mRAW getting a little softer, and full RAW was tack sharp.

Conclusion

I wasn't too thrilled with the savings/benefit with medium (mRAW), the difference in files size is only about 2mp less than a full RAW when converted to DNG.  Furthermore,  mRAW files bogged down my machine almost as much as full RAW did, and so why not just shoot full resolution anyway?

However, I find that its safe to say, at least for my testing, that sRAW is certainly suitable for both online use and sharp prints up to 8×10, and if carefully controlled, 11×14. How many images in your library fit within those guidelines?

Finally, it's up to you. like my friends whom shoot and keep everything, they are willing to deal with the additional hassle and added expense of archive and maitenance. Granted, much of it has to do with the nature of their work, and some of it has to do with the "You just never know what you'll need." as well. However, based on my findings, it's certainly a good argument to consider incorporating sRAW in your arsenal of smart image management.

Keep Shooting.