High-Definition, Deep-Focus Still Life (with Photoshop Auto-Align Layers and Auto-Blend Layers)

◆    ◆    ◆

– This article is intended mainly to explain in depth the techniques and procedures I’ve evolved, using mainly Photoshop’s Auto-Blend Layers and Auto-Align Layers commands, in order to make extra-large digital images which have at the same time a depth of focus that is not possible by ordinary means.  It is aimed at relatively serious photographers who are competent in Photoshop and who are willing to do something technically a bit ambitious (really more arduous than difficult), or at anyone who wants to know in detail (whatever be the reason) about such a procedure.

– It is not, however, a set of instructions – for this, see the article Step by Step Instructions, right nearby here in Artist's Statements.  (A shorter, more general article, "An Introductory Essay", grouped with the photos themselves, gives a simpler account of these procedures, as well as the background of their development, repeating some of the substance of this article.)  My purpose is somewhat didactic, so I will be mentioning aspects that some will find either elementary or tediously precise.  Along the way there are numerous digressions, mostly relegated to footnotes, indicated by bracketed, small-case Roman numerals, to be found at the bottom of the page.

– By “high definition” I mean extra-large images assembled from two or more separate photographs of different sections of a subject, thus achieving a higher definition (or a larger image size) than would be possible with a single image, which is of course limited to the pixel resolution of the camera used.  In my practice, this usually means four or six sections of a subject, moderately rectangular or square, in a 2x2 or 3x2 configuration.  At times, with a very long subject, just two sections are butted together in a 1x2 grid, the long way, or three sections in a 1x3 configuration, the short way.  (I once photographed a subject in twelve sections, but this is unmanageably big.)  A full frame from the Canon 5D is 5616 x 3744 pixels, which at 300 pixels per inch (a generous standard resolution for high-quality printing) is 47.55 x 31.7 cm, or roughly 18-3/4 x12-1/2 inches; multiply these by two or three to get an idea of the difference.

– By “deep focus” I mean not just the satisfactory depth-of-field any competent photographer knows how to control, but a total focus achieved by combining the sharpest parts of several separately-focused images of the same subject.  Think of a sandwich of different focuses, of which the sharpest areas will be eventually shown and the rest hidden.  If you want to visualize the overall structure of an image assembled this way, in terms both of multiple sections and deep focus, imagine (presuming a 2x2 configuration) four rectangular sandwiches butted up against one another: each sandwich is a section of the overall image, and the layers of the sandwich are the separate, differently-focused layers of each section.

– I make still-life photographs in the studio.  My subjects are generally smallish objects, either found or fashioned by me, arranged in boxes, trays, drawers, or shelves, sometimes compartmented, which serve both as settings and as compositional or organizing devices.  These “containers” are usually of small and medium sizes, from perhaps 6”x8” to 24”x30” or so, and rather shallow, say, from two to four inches in depth.  They are photographed straight-on.  Technically they could hardly be easier – when I made them with the view camera I made only the slightest movements of the swings and tilts of the lens or back (to adjust either the plane of sharp focus or the rectilinearity of the subject on the ground glass).  I keep the lighting simple, a large box light or tent, with electronic flash.  When I used a 4”x5” view camera for these photographs, I did quite well with the depth of focus using an aperture of f/22 or f/32 or so, using either a moderately wide-angle or moderately telephoto lens.

– Depth of focus is not ordinarily a problem when photographing frontally a relatively flat, shallow subject. In fact, a single exposure might have perfectly adequate depth of field, and it wouldn’t be worth the trouble of assembling a multiply-focused combination.  But it becomes problematic when photographing a subject in sections, because doing so means throwing the camera off-axis with respect to the subject as it would have been photographed head-on in a single frame.  (Meanwhile the camera is kept – on a sturdy tripod of course – in basically the same position.) There is an additional degree of depth that needs to be covered due to the oblique perspective.

– You might think, why not shift the camera left or right, up or down, with respect to the subject, instead of aiming it this way or that to frame a given section of the subject?  (Or shift the subject and keep the camera in position?) With a rectilinear subject you might even keep all your sections square to the frame.  And of course you could do this, and then assemble the sections with Auto-Align Layers.  But remember that perspective is entirely a function of the point of view, that is, the position of the camera in relation to the subject.  Shifting the camera (as opposed to changing your aim) is certain to give you noticeable discrepancies of perspective – parallax errors – in the adjoining areas of the sections where the same objects are seen.  Auto-Align Layers can sometimes “fudge” these transitions acceptably well by choosing one area or another where sections overlap, but it can’t correct perspective within a local area of an image – only whole sections.  It won’t join and smooth everything out for you; it isn’t that good.  (It can only do the same transformations you could do “by hand” with the Transform command, and do them simultaneously, but without making any purely local adjustments at the same time.)  If, on the other hand, you keep the camera at a fixed point and merely change your aim, the separate sections will be much closer to agreement in perspective.  You may have discrepancies unresolved by Auto-Align Layers, but they will be much easier to deal with than if you had shifted the camera from side to side and up and down. [i]

– With view camera and film (this digression is not relevant to the techniques that are the subject of thus article, but serves to suggest something about the aesthetic considerations I make in this work, and a little of the background of my experience), I photographed a few subjects consisting of two or three related panels, photographing first the entire ensemble, and then the separate panels.  (They are a bit reminiscent of medieval altarpieces composed of a number of panels; in fact I usually call them diptychs or triptychs.)  In making the images of the separate panels, I had the choice of simply aiming the camera left or right, or of moving the camera to one side or the other while keeping it oriented frontally, or of shifting only the camera back left or right in order to frame that part of the subject on the ground glass without moving the lens – thus perfectly maintaining the perspective within the set of images.  If you proceed in this last way, you will be able to view the entire printed piece from the same perspective you’d have if you were standing before the actual subject.

– This is a subtle difference, having mostly to do with the degree of the viewer’s feeling within the actual space of the subject.  It’s a bit like the difference between feeling physically closer to the subject of a wide-angle image, and more distant from one made with a longer lens. To maintain consistent perspective throughout the panels may smack of illusionism to some (who are by nature or on principle opposed to such things), but it does heighten one’s confrontation with the subject (in a way that is not obvious and vulgar like most 3-D imagery).  A portrait lens or telephoto lens is less intimate than a wide-angle, and tends to give a somewhat abstract perspective on the subject, which induces in turn a more abstract experience in the viewer (possibly even cerebral, if the necessary RAM and a good imagination card are there).  This may seem like hair-splitting to some people, but to me it counts for something.  In practice I usually photograph multi-panel subjects both ways, to have both a consistent perspective for a unified piece, and always-frontal perspectives for the separate panels to be seen (and sold) independently.

– Since for many years I did all my more serious still-life photography with traditional film and a 4”x5” view camera, it has been a disappointment to be constrained to switch over to digital work.  Above all this has been due to the limited resolution in pixel size possible with most digital cameras or backs – at least the ones that don’t cost many thousands of dollars – compared to the definition you get from large format film.  I have used a few cheaper digital cameras (moving up, every few years, to something more advanced), and have generally been quite satisfied with the color and tonal rendition of the photos, but not with the sharpness, which never compared with what I could get even from a traditional 35mm film and a good lens (let alone a view camera).  These digital photos also had an annoying chromatic aberration towards the corners of the frame, quite noticeable on examination, that I never got in the old analog days. [ii]

– This is largely the fault, I suppose, of the zoom lenses built into most of these lower and mid-range cameras – made for amateurs.  With traditional means, whether view camera or small format, I always used a lens of a certain quality, and so avoided most such problems.  Speaking of chromatic aberration, Lightroom does a fine job correcting it in RAW files, just by choosing “Remove Chromatic Aberration” (in Lens Corrections in the Develop module) – though it’s a nuisance to have to do it at all.  And – while we’re still in that little box – I’d advise against the routine use of “Enable Profile Corrections” to do basic corrections of perspective distortion, or the use of the manual corrections, if they can be avoided, since they reduce slightly the sharpness of the image.  With conventional, single image photographs, I use them only when necessary, e.g., when the barrel effect of a wide-angle image is disturbing, as with some rectilinear subjects, or if I happen to have taken the picture a little cockeyed.

– They are also necessary (I hasten to add) with the overall base images of my deep-focus, hi-def procedure, at least if there is any distortion or undesired tilting, and with the images of the separate sections as well if there is perspective distortion. (This will depend on the lens and the focal length used.) [iii] The procedure I am outlining here involves two unavoidable transformations (the initial one of Load Files in Series, which aligns a series of images with different focuses, thus adjusting their sizes, and the final one of Auto-Align Layers), both of which reduce definition by a couple of hairs.  These are extremely slight and tolerable, but I wouldn’t do them if they weren’t necessary.

– To get this, I had to resort to photographing my subjects in sections, and then reassembling them in Photoshop with Auto-Align Layers, using a single, overall image as a base.  Depending on the shape of the subject, this can mean as many as six sections.  A typical subject photographed in a single image measured some 4400 x 3700 pixels, but exposed in six sections, with a bit of overlapping, the final assembly measured almost 11,000 x 9000 pixels.  (In twelve sections, it came to around 16,000 x 13,500 pixels, but this was becoming unmanageably big.  And this was not just for the order of magnitude of the job, but for sheer file size – my Photoshop limits TIFF files to 4GB, so one is forced to keep the elements of the assembly under control.)

– In the studio I mainly do still life, often of rather small arrangements, and so I’m working at rather close range most of the time.  This means that depth of focus is always an issue.  To deal with this, I make a series of exposures with different points of focus, and then assemble them in a sort of Dagwood sandwich with File/Load Files in Series. Then I apply Modify/Auto-Blend Layers to get a good start on the masks that determine the sharpest parts of every layer.  Once these layer masks have been adjusted to my satisfaction – the longest part of the procedure – I make a copy of the file and unify the layers.  Then I combine the unified layers of all the sections in a new document, along with the single, overall shot (adjusted beforehand in Photoshop with Transform, if need be, to make it as rectilinear as possible).  The overall shot I enlarge enough to match the size of the separate sections – this is usually around 240% – and put it below all the other layers, locking it. I position all the layers as much as possible to conform to the underlying base layer.  Then I select all the layers, including the base, and do Auto-Align Layers.  Before going any further I examine the result very closely, looking for imperfections of alignment which may be edges that are do not follow the rectilinearity of the base, or simple failures to align perfectly within the image.  I evaluate what will be necessary to do “by hand” in Photoshop to perfect the image, often making little notes (squiggles, arrows, etc.) in a separate, new layer called Problems.  If it looks too difficult, I may go back in History to just before Align Layers, re-arrange the sections slightly, and try again.

– In practice I make a series of trial exposures of the whole subject using the 50mm lens.  (At least that’s what I used up to a point.  Now that I have a much better 24-70mm zoom lens, I use that, and make my overall exposures at a focal length of about 35-50mm.)  When I’m sure I have what I want, I make a series of definitive exposures as well, bracketing a bit.  Then I switch to the 100mm Macro lens, back off from the subject enough to include the whole arrangement again, and make a series of exposures again.  Finally I come in close again – actually to roughly the same distance I had with the shorter lens – and make the exposures of the separate sections. [iv] (The astute reader will realize immediately that this distancing changes the perspective of the image, and of course it does.  But I make the overall, 100mm exposures basically just to have them, as they are very easy to make.  I may even use them as an alternate version to the much larger, definitive photograph.  Still, I prefer the closer, more intimate perspective of the shorter focal length (replicated in the perspective of the combined sections), as opposed to the more distant feel of the longer-lens exposure.  As it happens, the close-ups of the sections have about the same perspective as the 50mm overall exposure, so the latter serves not only as a good visual guide to what I’ll get with my multi-section assembly, it also makes a good base for the Auto-Align Layers transformation.  The less ambiguity for Auto-Align Layers, the better.)

– My sincere compliments to Adobe for providing these very useful tools.  I look forward to future improvements.

Allen Schill

October 2013

© 2014 Allen Schill.  All rights reserved in all countries.  No part of this document may be reproduced or used in any form without prior written permission from the author.  Anyone is welcome to link to it or to quote brief passages, but I would like to be notified.

Link to relevant section of AllenSchill.com:  Deep Focus Still-Life Photographs


[i] Truly perfect agreement is possible if you determine the point of the true optical center of the lens, and change your aim from this central point, thus avoiding the slight change of perspective introduced when adjusting the aim by means of the tripod head.  (There are tripod heads dedicated to this purpose, sometimes called nodal point heads.)  This point – and here my understanding of optics gets shaky – is sometimes inaccurately called the optical center or the front nodal point, but more correctly the no-parallax point or the entrance pupil.

Just to contradict the point I have just made – occasionally these lateral shifts of the camera can work, provided the parallax errors will not be evident, or can be fudged in postproduction.  That is, in the case of aiming straight down at the subject, you may be able to shift the camera left and right, up and down, without going oblique.  The convenience of this method, presuming it is suitable for the subject, depends largely on the movements possible with the kind of tripod and tripod head that are used.  I use a column support for the camera, with a boom that extends to the left of the column, and a base attached to the boom on which is mounted the tripod head and the camera.  With this, it's very easy to shift the camera left or right, quite precisely – I just crank the boom in or out. But no such easy movement is possible for going forward and back – I can only release the brake on the tripod base, slide the whole column forward or back along the floor, and put in the brake again.  This is very awkward and imprecise; if I manage to get it right in only ten attempts of humping, pushing, and looking, I am content with my luck.  Honestly, I try to avoid this method, which is truly easy only with subjects that can be covered with two or three lateral sets of exposures.   (The boom itself has limited extension.)

Due to these eventual difficulties, it may be easier and preferable to shift the subject left, right, up, and down, while keeping the camera stationary.  The trouble with this, however, other than the delicate task of moving the subject in a precise way, is that the illumination may not agree from one section to another, because – relatively speaking – the position of the light changes with respect to the subject; it does not change when you shift only the camera.  You can get away with it with fairly small subjects without too much depth; I have done so in a few cases – e.g., the "Untitled (Fagiolini)" and the "Untitled (Pistachio Plaster Relief".)  The big advantage of either of these methods is that depth of focus is much less of a problem than when the camera is aimed obliquely at the subject.  As such, far fewer exposures are needed to "cover" the depth of the subject.

[ii] To be fair, though, I should get out the light box, find the loupe, and look carefully at a few old slides to see if there was some aberration I simply never noticed.  But it’s hard to believe that I might’ve missed such a degree of color fringing in all these years.  For all I know about lens design, it may be that lenses for digital cameras are more inclined to this defect than lenses made for use with film, though this seems improbable.  More likely it is instead a matter of general lens quality, and/or a tendency with zoom lenses.)

[iii] I don’t know why Adobe – and a few tutorials I’ve seen – don’t warn the user of this, although the fact will be obvious to more experienced Photoshoppers, and evident – though perhaps just barely – to anyone who observes the before and after closely (at 100% or 200%) in Lightroom.

[iv] Regarding these lenses, I was surprised to learn that the focusing ring does not stop at infinity – which I only discovered when I chanced to take a few landscape photos, and saw the image of a distant subject go into focus and then out again at the extremity.  I was further stupefied to learn from Canon that this is normal, that this is the way these lenses are made – something to do with the design of autofocus lenses, which must mechanically allow the over-focusing in order to function as autofocus.  I still don’t quite understand why this should be necessary, but sometimes you just have to take things on faith.  My surprise was due largely to my inexperience with such lenses, which perhaps have always been like this.  But it took away a major convenience of taking photographs at a distance – I could simply turn the lens all the way to the end and be quite sure of good focus.  (That’s pretty automatic, I think.)

© Copyright Allen Schill

Using Format