Friday, January 20, 2023

Post Processing for Deep, Wide & High Dynamic Landscapes

In the last post I discussed a few rules of thumb that I use to capture wide angle landscape images, that require multiple brackets for focus and dynamic range. 

Rather than capture exposure brackets at each focus station, my prefered approach is to focus bracket for the foreground at a single exposure and then, when I'm focused at infinity, capture one or two exposures for the sky.

In this short post I'll discuss how I Post process the resultant bracket set.

The following is a test capture that I did in my garden. I used a Canon M3, at an 11mm focal length, with an f/6.3 aperture, at ISO 100. As I was using my M3, I also used my auto bracketing Lua script to capture the required image set from the minimum focus distance through to infinity:

As can be seen, the script took seven focus brackets for the foreground at 1/15s, out to around three times the hyperfocal, ie infinity; and one infinity exposure for the sky at 1/100s.

The post processing workflow that I use as my base approach, eg I won't necessarily follow every step, every time, goes like this:

  • Ingest the image set into lightroom
  • Pre-porocess with  DXO PureRaw 2 (or another RAW processor, eg Lightroom) if required
  • Adjust one of the foreground images, ready for focus bracketing, and sync with the other foreground images
  • Export the foreground images to your focus stacking software, eg Helicon Focus, Zerene Stacker or Photoshop
  • Focus blend the foreground images and, if required, export to Photoshop
  • Back in Lightroom, adjust the sky bracket ready for exposure blending with the focus stacked foreground image
  • Export to Photoshop
  • Bring the focus stacked forground image and the sky image together into a two layer document
  • Align the images if required
  • Adjust the exposure of the two images if required, ie so they will blend together well
  • Place a simple gradient mask on one of the images and increase the feathering of the mask to achieve a perfect blend. That is, there is no need to play around with luminosity masks etc
  • Flattern and clean up the image as required, eg cloning etc
  • Return to Lightroom and finish off the image as required, eg toning, dodging and burning
  • Process with Topaz Photo AI (or another such app or in Lightroom) to ensure noise and sharpness is covered, once again, if required

The resultant focus and exposure blended image, with a little bit of toning, looks like this. In this case I did all the focus stacking and exposure blending in Photoshop. Note the image is far from complete ;-)

Hopefully this post, together with the previous posts, has convinced the reader that capturing and processing deep focus and high dynamic range scenes is easy: at least with a wide angle lens.

As usual I welcome any comments on this post or any of my posts.

Sunday, January 15, 2023

Rules of Thumb and capture Workflow for Wide Angle Landscape Photography

In previous posts I've discussed various models and rules of thumb for wide angle landscape photography. In this post I'll layout the workflow I personally use.

First, let's remind ourselves of a few rules of thumb that cover both focusing and exposure:

  • The hyperfocal distance,(H) in meters, may be approximated by dividing the lens focal length (in mm) by 10. This assumes the aperture (N) is set to f/10 and you are content that the circle of confusion (C) is the focal length taken as microns. To find the hyperfocal at another N or C simply use the following expression to adjust the F/10 value:
  • If the lens throw between the nearest point of focus (x), as measured from the front principal, and infinity is T, then this should be divided by H/(2x) to derive the number of focus brackets;
  • A reasonable estimate for the position of the front prinicpal of a wide angle lens, eg when doing landscape photography, is to assume it is at the non-parallax point of the lens, ie the entrance pupil, which may be esimated in the field in realtime, if not known in advance.
  • For high dynamic range scenes, ie where you are not confident you can capture both the sky's highlights and the foreground's shadow details in a single exposure, set an ETTR exposure for the sky and adjust the exposure for the foreground by +4Ev, ie four stops. Together with noise reduction software, such as Topaz PhotoAI, a +4Ev approach will cover most/all situations. This allows us to take a fixed exposure when focus bracketing for the foreground, and an ETTR exposure for the sky, ie at infinity.
All the above rules of thumb are designed to be 'calculable' in your head. The accuracy of the rules are appropiate for wide angle lens use, but not macro lens use. For example, by assuming the front principlal is located at the entrance pupil, the most we will be out is F*(1-1/p), where p is the pupil magnification, which for most wide angle lenses, ie retro focus lenses, will be greater than unity. Thus, even if p is high, the most we will be out will always be less than the focal length of the length.

As for a capture workflow here is the one I follow:
  • After visualising and composing the scene, I assess the dynamic range of the sence and decide if I can 'get away' with a single exposure capture throughout, or if I need to adopt a +4Ev approach, ie ETTR for the sky and adjust the exposure for the land by up to 4Ev. If the DR is greater that 4Ev, then consider two exposures for the sky, at the ETTR and one between the ETTR value and the exposure value for the land;
  • Estimate the hyperfocal and if the nearest object of interest (x) is closer (to the entrance pupil) than half of this, then prepare to take some focus brackets. If not then focus slightly beyond the hyperfocal by, say, at least 2*H and take a single focus capture (with or without an ETTR as required);
  • If the number of focus brackets from H/(2x) is less than say,  5, divide the lens throw (T) by that number and take the indicated focus brackets; for example in the case of H/(2x) being 4, at the quarter throw points, ie at the nearest focus point and at 1/4, 1/2 and 3/4 of the throw and, finally, at infinity - giving 5 focus brackets in total;
  • If the estimated number of focus brackets are more than 4, then assess if it’s still ok to just take 4 focus brackets. For example, if the C is, say, 12, ie I’m using a 12mm lens, then I can afford to double the bracketing C to 24 microns, assuming a full frame camera, thus, in this case, I can still divide the throw by 4, even if H/(2x) is up to 8. But note that the acceptable value for C will depend on the size and how you display your image, eg social media vs a print in a competition;
  • Capture the focus brackets at the exposure for the ground and, when at infinity, adjust the exposure for the sky and take the sky capture(s).
Although the above may appear complicated, I can assure you it is not and, after practicing, you will soon feel at ease with the workflow.

I’ll discuss post processing in a future post. 
As usual I welcome any comments on this post or any of my posts.

Friday, January 6, 2023

A Rule of Thumb approach for perfect landscape focus bracketing with any (WA) lens/camera combination

Those that read my blog know I love 'playing around' with focus stuff. I'm always looking for hacks to help me, pragmatically, use all my gear: both AF and manual lenses.

In the last few posts I've shown how we can use a split, thin lens model, together with an estimate of the position of the front principal, to create any required depth of field scale.

In this post I will show how we can focus bracket by exploiting the lens itself. So, there is no more need to follow incorrect, urban myths, eg focus a third into the scene.

First, let’s remind ourselves of the Rule of Ten, that gives us an estimate of the position of the hyperfocal from the front principal.

The hyperfocal RoT says that at f/10 and at a circle of confusion of f, as microns, the hyperfocal distance, in meters, is simply the focal length of the lens divided by 10.

As an example, a lens with a focal length of 16mm, set to f/10, will have a hyperfocal at 16/10 = 1.6m, using a CoC of 16 microns.

The hyperfocal RoT is therefore ideal for wide angle lenses, ie where the focal length is less than the largest acceptable CoC in microns; however, it is simple to use any lens length or CoC by adopting the following form of the RoT, where the terms in the square brackets are optional:

H = (F/10)[*(10/N)*(F/C)]

As an example, take a 50mm lens at N=10. Without adjusting things, the H would be 5m, but at a CoC of 50 microns! To adjust to a more sensible CoC, say, 25 microns, all we need to do is adjust the base H by (F/C), ie 50/25 = 2. Thus, the hyperfocal, at 25 microns, is 10m.

Once we have an approximation for the hyperfocal, we can now make use of hyperfocal bracketing:

From the above we can see an estimate of the number of brackets from the current focus (x, as measured from the front principal) to H/2 is simply H/(2*x): which we should round up.

For example, if we have a 16mm lens set at f/10, the RoT hyperfocal is 1.6m at a CoC of 16 microns.

If we now have a near point of focus at, say, 200mm from the front principal, we can estimate the number of focus brackets from x to H/2 as 1.6/(2*0.2) = 4.

From this all we need to do is to exploit the lens helicoid to give us our perfect focus bracketing. That is, divide the lens throw, ie rotation angle, between the near point of focus and infinity by 4; rather than guess where to focus.

Thus we end up with 5 focus brackets: four covering from the minimum point of focus to the H/2 point of focus, with a far depth of field at H; and one at infinity, ie with a near depth of field at H. 

In lens rotation space things look like this, where we see the two end focus points, ie at the near point of focus and infinity; and the three other points of focus, positioned at the quarter throw locations:

Clearly, once the number of brackets starts going beyond, say, 4 or 5, things get a little out of hand. However, the above hack, exploiting the lens throw/helicoid, is obviously better than guessing where to focus; and, on a wide angle lens, you will nearly always only need a few focus brackets to ensure deep focus capture.

As usual I welcome any comments on this post or any of my posts.

Wednesday, January 4, 2023

Focus bracketing with the AstrHori 18mm F8 Full-frame Wide-angle Shift Lens

By way of an experiment and a bit fun, I thought I would treat myself to a rather bulky ‘caplens’ for my EOS R. 

Rather than buy something like the Funleader 18mm F8.0 CAPLENS, I decided to go with the slightly larger AstrHori 18mm F8 Full-frame Wide-angle Shift Lens. Yes a shift lens:


In the above we see the focus scale and the +/- 6mm of shift.

As this is a cheap lens, there are two weakness that need to be recognised.

The first is only visible when you attempt to use the  maximum 12mm of shift: +/- 6mm. Clearly the lens elements are only designed for an APS-C image:

But at around £100, not unexpected.

The other weakness is clear if you look at the lens focus ring: there is no depth of field scale and this is a fully manual lens, with no electronic connection to the camera.

But, as we have seen in the last few posts, we can easily construct our own depth of field scale and, in fact, can create it in a rather interesting way, as the lens has a fixed aperture, which allows us to construct the DoF scale with respect to blur in microns, ie circles of confusion. 

Using the model described in the last post, the DoF scale can be constructed from d = ((MFD-X)*L*N*C)/(F*F), and calculating d for Cs at 10, 20 and 30 microns. Giving, for the 18mm AstrHori, ds of 4.5mm, 9mm and 13.5mm, ie for the near and far DoFs at CoCs of 10, 20 and 30 microns. 

From L is 57mm, F = 18mm, N = 8, the MFD is 340 and X (the estimated position of the front principal, relative to the sensor plane) = 30mm.

Marking out the scale on some masking tape results in the following DoF scale hack:

Using the scale is no different to using an aperture based depth of field scale. Namely, select the DoF criterion you wish to use, ie the CoC. In the example below I used 20 microns, focus on the nearest object of interest,  the minium focus in this case, capture an image, and rotate the focus, using the 20 micron marks, taking images and repeating, until you reach infinity.

The following is a four focus bracket test capture set I grabbed on my dining table:

So, there you have it, a cheap shift lens that can be made better with a bit of masking tape.

As usual I welcome any comments on this post, or any of my posts.

Saturday, December 31, 2022

Making pragmatic sense of the Depth of Field scales on manual prime lenses

In previous posts I have discussed how we can use a split, thin lens approximation (image below) to create a depth of field scale for any lens, at a given focal length, aperture and assumed circle of confusion.

The expression for the ‘length’ of the near and far depth of field scale (d=J/h) being: ((MFD-X)*L*N*C)/(F*F). Where: MFD is the minimum focus distance, as measured from the sensor; X is the position of the front principal, as measured from the sensor; L is the ‘rotational distance’ travelled, in mm, that is the lens throw, as measured from the minimum focus to infinity; N is the aperture number; C the circle of confusion; and F the focal length of the lens. 

In this post I’ll show how we can apply the model to manual prime lenses, where there is usually a very good depth of field scale on the lens, but we don’t know what circle of confusion (C) was used to create the scale.

Knowing the value of C is useful for landscape photographers, when: hyperfocal focusing to a specific CoC; setting a specific infinity blur; or ensuring a specific overlap blur when focus bracketing.

To find an estimate of the DoF scale’s value of C, we simply rearrange the expression for d, to obtain the equation for C. Namely C = (d*F*F)/((MFD-X)*L*N).

As discussed in a past post, we can estimate the value of X, the distance from the sensor plane to the front principal, by looking up the value, say, on Photons To Photos, or, if PTP lens information isn’t available, then we can estimate X by locating the entrance pupil, eg by using parallax, and measuring/guesstimating the pupil magnification.

As for d and L, these can be directly measured from the lens by placing a strip of masking tape on the lens and marking the location of the MFD, ie 0, the infinity location, oo, and the length of the depth of field scale at the maximum depth of field location, eg f/22.

The following image shows three measurement records, for a Mamiya 645 35mm Sekor lens, the Irix 11mm, and the Pergear 14mm: ignore the faint text as this is print through from another page.

In the above the MFD is at 0, the infinity location on the far right, with the length of the near to far depth of field shown at 22 or 16, with the left hand (near) DoF being placed at 0. Thus L is the length from 0 to oo and d is half the length of the 0 to f/22 or f/16 length.

Having got a model and a process to gather all the data, let’s look at a these lenses.

The first lenses we will look at is the Mamiya 645 lens, which I personally mount on my EOS R via a Rhinocam Vertex adapter, to create a quasi medium format camera, ie using sensor bracketing to create a ‘square 645’ image, equivalent to a sensor of about 45mm square, eg about 8500x8500 on the R.

In the case of  my Mamiya 645 35mm, on the R, the measured MFD = 420mm, with L = 110mm, d = 29mm and X estimated at 80mm. Thus the circle of confusion used for the depth of field is about 43 microns.

Now some may be saying, so what, I didn’t need to know this to use the depth of field scale. The manufacturer did all the work for me. But, as we will see, you can’t always trust the manufacturer, or, put another way, don’t assume the CoC: always confirm the DoF scale’s CoC; especially if you are seeking to set focus to a specific optical blur.

To illustrate this, let’s look at the other two manual prime lens.

First, the Irix 11mm Blackstone, where the MFD is 270mm, with L = 100mm, d = 18mm, an estimated (but not confirmed) X of 90mm; giving the circle of confusion used for the depth of field on the Irix at about 8 microns. That is not the ‘standard’ CoC that some may assume for a full frame camera, ie around 30 microns.

Finally, let’s look at the Pergear 14mm using an MFD of 430mm, with L = 53mm, d = 16.5mm, an estimated X of 60mm; the circle of confusion used for the depth of field on the Pergear 14mm is also about 8 microns. 

So what does all this mean?

First, it looks like there is a difference between the older, film lens depth of field scale and the two, new digital lenses. That is the Mamiya appears to use a ‘standard’ CoC for the 645 format, ie around 43 microns; whereas the two modern lenses seem to use a CoC much less than the 'format, standard CoC' of around 30 microns, ie instead around 8 microns.

Taking the Mamiya as an example, and knowing the CoC that has been used by the manufacturer to construct the depth of field scale, we can now use this information to ‘adjust’ focus settings in the field.

For example, if I was focus bracketing with the quasi medium format setup, using the Mamiya, I would be seeking an overlap blur that is, say, half to 2/3 of the standard format CoC. Thus, if my aperture is f/16, then I would use the f/8 depth of field markings to focus bracket, knowing this will result in a (medium format) overlap blur of around 20 microns. 

In a similar way, if I wish to set the infinity blur to, say, 20 microns, once again I would place the infinity against the f/8 mark, with an aperture set at f/16.

However, on the Irix and Pergear lenses, I would adopt a different strategy, knowing the CoC that has been used to construct the depth of field scale is around 8 microns.

All this assumes I’m using the Mamiya lens as a film camera, but I’m not. Hence, knowing the value of C that has been used, ie 43 microns, on my R I might be tempted to use an aperture of, say, f/16, but use the f/4 depth of field mark. Having said this, I think there are a few more experiments I need to make, before I’m confident about things ;-)

In future posts I’ll explore the above with further examples, but, for now, I'll simply finish this post here.

As usual I welcome any comments on this post or any of my posts.

Sunday, December 18, 2022

DoF Hack: the pretty version

In a previous post I showed how one can easily create customised depth of field scales on any lens, eg when wanting to focus stack from near to far, say, from the minimum focus to infinity. The hack allows you to choose a particular circle of confusion, rather than accept the one 'baked into' the lens by the manufacturer.

The reason for wishing to create such a hack is if the lens (manual or AF) doesn't have an apprroiate DoF scale, or the electronic feedback in the camera, once again, doesn't provide the necessary infomation, ie how much to rotate the lens helicoid, to align the previous near DoF with the previous far DoF.

In the original post I showed the basic hack using trusty duct tape:


It's functionally good enough, but it hardly looks pretty or professional ;-)

Luckily my wife has a Brother P-Touch label printer and, as can be seen below, using this you can create a much prettier layout:

As before, to use the DoF hack, all you need to do, according to the direction of focus bracketing, is to rotate the lens focus ring (on the Sigma above there are no hard stops, only soft stops), until the 0, say, aligns with the minimum focus distance: or align the 9 with infinity.  

BTW the DoF above is for a 12mm focal length, at f/8 and with a CoC of 20 microns. All you need to do is to cut the appropriate DoF length, based on focal length, aperture and CoC, using the technique discussed in the first post.

The numbers on the tape are only there to assist in your rotation indexing, eg from the left to the right of the DoF bar.

As usual I welcome any comments on this post, or any of my posts.


Friday, November 25, 2022

Deep Focus Bracketing Hack: Addendum

In the last post I discussed a depth of field hack that can easily be used to help manually capture perfect deep focus, landscape stacks, ie when non-macro focus bracketing.

The hack was based on knowing the location of the (infinity) front principal from the PhotonsToPhotos (PTP) Optical Bench Hub, and ignoring any lens extension, as the hack is aimed at non-macro, wide angle photography, where the magnification and lens extension impact will be small.

But what if your lens isn’t in the PTP database?

In this post I show a quick approach to estimating the front principal location, from a single, simple measurement.

Although it is difficult/’impossible’ to directly measure the front principal (H), it is relatively easy to measure the location of the on-axis entrance pupil (EP) location, P in the diagram below. At least for a ‘normal’, non fisheye lens.

In the above example, of the Sigma 12-24, at 12mm, we can see (by observing the ratio of the exit to entrance pupils), that the lens is, as expected, a retro focus one, with a pupil magnification (p = Exit Diameter divided by Entrance Diameter) of about 6. Thus, using the thick lens  model, we know that EP is offset, ie in front of the front principal H, by F(1-1/p): about 10mm at a focal length of 12mm.

The value of p can be measured, eg taking an image of the lens from each end, or ‘eyeballed’ at a reasonable level of accuracy.

As was shown in a previous post, estimating the EP, which is the pano rotation location, can be accomplished very simply using a light source. At a push, one could also, whilst on a tripod, locate the pano pivot location and measure the EP from the sensor plane directly.

Whatever approach you use, by estimating/measuring the position of the EP you can get an estimate of the location of the front principal from knowing the pupil magnification (p), ie the front principal is located at EP - F(1-1/p). Thus, even if you don’t have a model of the lens, you can consider the lens as a black box and estimate the location of the front principal, by measuring the location of the entrance pupil and measuring/guesstimating the pupil magnification.

As usual I welcome any feedback on this post or any of my posts.