Depth of field and Sharpness: Conventional Thinking is Unrealistic Going Forward, Focus Stacking is One Solution
Depth of field for an image refers to the near/far range of “acceptably sharp” detail in an image, the idea of “acceptable” being entirely context-dependent: what is “acceptable” for a magazine-size print or web image may look blurry when displayed at large size, such as with a 4K or 8K television, or in conventional parlance, a “large print”.
When choosing a camera and making images, consider that the future is “large”: 4K now, and soon 8K and then presumably 16K displays with unprecedented dynamic range and extreme resolution capabilities (4K is 8.2 megapixels, 8K is 33 megapixels, 16K is 132 megapixels). Already, Netflix 4K streamed video reveals the slightest technical errors in spite of being compressed substantially: errors in focus, one eye sharp and the other blurred, soft eyes but sharp nose or ears, strictly limited depth of field, lens aberrations and not so pleasing bokeh effects.
Acceptably sharp also depends on subject matter: an image may have prominent structural features which may make fine details a secondary concern, or it might contain finely detailed elements, like hair, grass, pine needles, metallic textures and so on. A sense of realism vanishes when detail is lacking—the “transparency” of the image is impaired, even if the by-the-numbers resolution is good by conventional standards. Does the image look like the real thing or does it look like a picture of it? That is why I consider the late iMac 5K display the best display device ever created: it looks more like the real thing than anything else. But much larger screens with more resolution are only a few years off. [See What’s the Best Way to Enjoy Images at their Finest? and iMac 5K for Stunning Black and White Images ].
Traditional depth of field thinking is a concept ill-suited to modern display devices, because those devices are already really big (“big print”) and because they are transmissive devices that do not dither or degrade the pixels as do printed media and projection: every pixel can be rendered to perfection.
Cameras that in 2016 just barely fill the screen on mainstream computers like the Apple iMac 5K will barely fill half the screen within a few years. An Apple iMac 5K has a 5120-pixel-wide display that is dazzling for images whose resolution fills those 14+ million pixels. Similarly the 8+ megapixels of 4K television shows every weakness and error in focusing and depth of field. But 8K televisions and computer displays are not many years off, indeed 8K televisions already exist (8K is 33 megapixels). An iMac 6K or iMac 8K in a larger screen size is a distinct possibility (or some equivalent product or TV).
I consider 16 or 24 megapixel cameras somewhat a waste of my time; it takes the same effort to shoot 16 or 24 or 36 or 42 or 50 megapixels. Why would I waste my time at 16 or 24? For perspective, view circa 2000-2005 digital images of 3 or 4 or 6 megapixels on an iMac 5K at full screen. The 16/20/24 megapixel images of today aren’t going to even fill an 8K display; they will be “blurred up” to fit. They will be satisfactory of course, but not eye popping.
And so, traditional film-based thinking about what is “acceptably sharp” is already obsolete in relation to even current display technologies, let alone future ones. My view therefore is that “acceptable” depth of field is that which approaches the full sensor resolution, thus capturing an image in the most possible detail.
Unconstrained or misplaced depth of field can be a negative
More depth of field willy-nilly is not always better. The zone of sharpness requires careful placement for starters. And sometimes an image is better done tack-sharp in a relatively narrow zone, while still including background and/or foreground blur for separation of the subject from its surroundings.
Yet even where some blur for subjection separation is a goal (a common one), there may be practical constraints on depth of field with traditional approaches, particularly with higher magnification images. Hence new working techniques may be needed, particularly as sensor resolution gains are made. One such technique is focus stacking.
Stopping down versus focus stacking
The traditional approach for more depth of field is to stop down the lens, say to f/8 or f/11 or f/16. But with high resolution digital cameras, even f/11 is often insufficient depth of field, and the image as a whole quickly declines in quality past f/8 due to the dulling effects of diffraction. That figure will drop to f/6.3 or so for higher resolution cameras no later than 2017 (it is already f/5.6 for APS-C).
The traditional “stop down” approach is often sufficient for high sharpness over a desired range. But what about when even f/11 or f/16 won’t suffice and/or when high brilliance is desired (that is to avoid the dulling effects of diffraction)? The answer is focus stacking.
Depth of field as defined above (full sensor resolution) is extremely limited on high performance cameras, even at f/8 or f/11. By shooting multiple images, each at a different focus position, the images can be combined for limitless depth of field: focus stacking.
Focus stacking, handheld at 21mm, f/5.6
Here, I wanted a sharp foreground the pyramidal rock to be sharp, and the background also to be tack sharp. But particular subject had always defeated me on a depth of field basis, even at f/13—there was just no way to make the whole scene sharp. A tilt lens would not help here, since the tilt would cut through the vertical pyramidal rock and it is a key element.
Use of a tripod for focus stacking is all but essential for avoiding stacking issues, but here with the 30+ mph wind throwing me off balance, I could not even frame the subject accurately, let alone use a tripod, which would have vibrated strongly from the wind—handheld was the only option.
I wondered: could I focus stack with just two handheld shots for acceptable results? The composite image had some stacking artifacts, but with a bit of cloning to fix a few defects, this 2-frame stack at f/5.6 (!) delivers depth of field from the nearby foreground to the distant mountains.
Why don’t all cameras offer built-in focus stacking? The camera could do it extremely well (even handheld at reasonably fast shutter speeds), and in raw. A few cameras offer focus bracketing, at least one does JPEG focus stacking, but AFAIK no camera does it fully.
2 frames: first frame focused just in front of pyramidal boulder, 2nd frame where hillside slopes away to distance.
Timothy R writes:
I just read your post, it’s great. In this past year I’ve personally done two things; stitched panoramas (using a Really Right Stuff Pano Head), and stacked images; both for the reasons you mention – more pixels and precise control of the zone of exact focus. To my mind there is no other way, on my iMac and without much practice I’m able to create immersive images, perhaps not interesting … but the technique has to be learned and practiced in any case.
For stacking I use Helicon Focus and I have observed that different stacking algorithms produce different results. It’s not so easy to know which is best, however it would not surprise me to find such features in a Sony camera soon. If Sony’s A7 developer’s API would allow control of the focal position (so far it does not) then I would write an app to do the stacking myself, including all necessary calculations and, if possible, the post processing too.
Still, the same limitations as sensor shifting apply, so its limited in its application. However, in camera focus stacking would really speed up certain kinds of photography.
DIGLLOYD: the limitation on motion for focus stacking is different than pixel shift imaging in that motion is often easily dealt with by retouching (right inside Zerene Stacker and/or in Photoshop, basically cloning in one of the frames). As per above, there was a LOT of motion including me moving the camera between frames and a violent wind that was moving the brush and water. Yet the image succeeds.
With pixel shift, “motion” means both subject motion and any change in lighting intensity or color; in either situation the issue becomes an ugly grid artifact at the pixel level, which is much harder to deal with—it requires full raw converter support. Pixel shift would be impossible in the above image, even for a single frame.
Bruce B writes:
Version 4 of the firmware for the Olympus EM-1 includes a focus bracketing mode and a built in focus stacking process. Internal focus stacking is limited to JPEG and has some limitations on number of frames, focus step size, etc. Focus bracketing is far more flexible, working with raw images and allowing a wide range of control over focus step size and number of focus steps.
DIGLLOYD: ironic that auto focus stacking is implemented in a camera whose format in which the feature is least useful (lots of depth of field with Micro Four Thirds). And JPEG is obviously not so desirable. Still, it’s a good idea that can be taken further. But the highest value accrues to the largest format (35mm or medium format), where depth of field is harder to come by.
The ideal camera would offer auto-stepping between two points of focus (distances) that is suitable for the chosen aperture. Then the camera could do all the work of dealing with focus. On the other hand, manual focus lenses are highly desirable for control. A camera ought to offer a “focus stack this and the next or last N images” feature, delivering a raw output file. Nikon has long had a multiple exposure mode that outputs a raw; this would just extend the idea.
Lawrence F writes:
The Phase One XF camera body has a focus stacking feature set in which the front and back focus planes are set and the camera automatically focuses and shoots between those set points. The photographer can determine how many exposures will be taken in that interval.
DIGLLOYD: I wonder if it does the right thing: fixed spacing is not appropriate for medium/far work. Also AF accuracy can cause spacing errors if things up if focused when stopped down, as I found with the Sony A7R II. Though the XF should not have that issue since it is a DSLR.
Christopher P writes:
Am I missing something here?
The most cursory glance at the Worlds’ most effecting and enduring images indicates that sharpness, let alone the ability to fill or challenge an 8K monitor, are secondary to content, form and impact. We have all seen any number of high resolution, well exposed, razor sharp images which, after a brief nod to the technical ability implied, are forgotten and never looked at again.
I own some of the high pixel count cameras and state-of-the-art lenses you espouse but, when I look at my images, taken over many years, my favourites are often those which look tiny on my 4K monitors.
A 5 megapixel camera in the right hands is capable of producing sensational images which make exquisite small prints.
By advocating that 16 and 24 megapixel cameras are a waste of your time are you not decrying the majority of the photographic community? I value the ability and wherewithal to take technically, exemplary photographs should the content demand but, more often than not, that ability is not essential, indeed it may even be deleterious.
DIGLLOYD: the response is a non-sequitur: I like chocolate ice cream, which does not mean I dislike vanilla. I like depth of field sometimes, which does not imply always. Indeed, focus stacking is appropriate for increasing depth of field but only in a narrow zone (via a “short stack” at wide aperture).
Nor is it necessary to conflate pictorial excellence (composition, content and timing and meaning) with sharpness.
More resolution rarely makes a photo worse, so it is indeed a waste of time to shoot a low-res camera that won’t even fill the screen when I can shoot a high-res one that will with no difference in effort, and therefore make an image that has greater transparency and realism. Anyone working with contentional resolution displays is working blindly and won’t be aware of the tremendous difference in visual impact that occurs with some images. That is true not just for sharpness, but for dynamic range and tonality, all parts of a whole. A great photo can be pure crap in a technical sense, but that is a non-sequitur to this whole depth of field discussion. For that matter a great photo is not subject to a vote unless one figures polls as a measure of greatness.
Finally, I have long advocated appropriateness in a wide variety of ways. See Landscape Photography at Wide Apertures and Aperture Series: Pine Creek Thunderstorms not to mention dozens of blog posts on those and many other topics. I am not advocating the most depth of field possible for every image, only those for which it is appropriate.