0

How can I calculate the depth of focus (Depth of field, DoF) for video recording?

There are some tools around that calculate the DoF for different camera types depending on focal length, f-stop and subject distance, e.g.

My assumption is, that the DoF for video recording is bigger than that for photographs, because video is recorded with lower resolution, and therefore a higher degree of unsharpness can be accepted without visible effect.

My idea was to calculate the size of a single pixel and use this size as "circle of confusion" parameter.

My question: Are these assumptions and the example calculation below correct? What is a reasonable way to calculate DoF for video?

Example:

  • Canon EOS 600D / Rebel T3i
  • f2.8
  • 16mm
  • object distance 2m
  • recording format 1080p

From the 600D's sensor size (22.3 mm x 14.8 mm) and video resolution (1920x1080) I calculate the pixel size to be 0.0116 mm x 0.0116 mm, or 0.016 mm in diagonal.

Entering these numbers into dofmaster, I get 1.6 m as a result (1.48 m ... 3.08 m).

What surprises me is, that using the camera model directly instead of the CoC, the result is 2.02m, which is bigger than that vor video.

Accordingly, when calculating for a video resolution of 720x576, I get a DoF of 146m.

  • I believe that many DSLRs do line skipping, which means that the pixel size is the same as when shooting a still, there's just more space between them. I'm not sure how that figures into the calculation, though. – user1118321 Dec 10 '14 at 14:46
  • I assume that both methods would yield almost the same results, apart from "non-line-skipping" functioning as a kind of low pass filter, and thereby reducing nasty Moire effects. I assume skipping or not skipping has no effect on sharpness impression. – nwsp Dec 12 '14 at 9:08
1

Well, your calculations appear to be correct, but the only calculation that isn't exactly known is the pixel size of your camera sensor. Either your calculation of the camera's pixel size on the sensor (at least when filming 1080 video) is off, or their recorded circle of confusion of that sensor is off. Or both. Though I find it more likely that their stored CoC value for that sensor is based on what is "reasonable focus loss" not on the actual pixel size, and so they have a larger CoC value than what you consider to be reasonable, as your reasonable is that the CoC is no bigger than a pixel.

Either way, calculating the differences based on resolution shouldn't be an issue. In the spirit of better safe than sorry, use your own calculation of the pixel sizes (0.016mm with the DoF result ~1.31m) as a basis. If you're going to test for half the resolution of 1080 (being 540) then you'll double the pixel size calculation (0.032mm) resulting in ~4m DoF.

To calculate for 720x576 (which is 576 resolution, not 720 resolution. Careful with that) you'd do (1080/576)*0.013mm = 0.03 (new CoC for calculation resulting in 3.44m DoF).

I'm not sure how you got 146m as a result though. Perhaps you input the wrong circle of confusion value?

0

Your impression is incorrect. Depth of field has nothing to do with resolution and everything to do with sensor size. Depth of field does depend on viewing distance, but that is a little different from resolution and not particularly applicable in this case.

For most online calculators that don't include a viewing distance, they are often assuming viewing the entire scene at once and that wouldn't make a big difference if it is a 2MP video image (particularly since the meaningful resolution of video is much higher due to blending) vs a DSLR.

  • I think this is true only as long as your resolution is higher than actually needed (which is normally true for photographs). Not so for video. Video resolution can be quite small and becomes a limiting factor. What can be regarded as reasonably sharp in 720x576, doesn't necessarily look sharp in full HD, being equivalent to a greater DoF at lower resolutions. Besides that - what do you mean with "is much higher due to blending" ? – nwsp Dec 10 '14 at 15:56
  • @nwsp - due to playing multiple frames, the effective resolution is much higher due to aliasing. When you look at a still from video, there isn't much information there, but when you watch a video, you aren't looking at one still. You are looking at a progression of related images and you pull information from more than one frame at a time. This is (one reason) why video often looks sharper when it is playing than when you pause it. – AJ Henderson Dec 10 '14 at 16:01
  • And you are correct that if you get close enough that pixelization is a problem, then you are going to have an impact on what looks blurry, but at that point the entire image is blurry, not because of DoF, but because of lack of resolution. Each piece of data you need is blurred with other nearby pieces of data. – AJ Henderson Dec 10 '14 at 16:02
  • And that's exactly the point. To which degree can I open the aperture without adding apparent additional unsharpness in the case of low (video) resolutions. Looking at erik-krause.de , the minimum resolution for "normal viewing distance" is always 5.7MP, way beyond normal (low quality) video. Hmm, maybe I should just play around with viewing distance until the computed minimum resolution fits the target resolution? Which results in 1:1.8 (image size : viewing distance) for 2K material, and 1:4.6 for 720x576 compared to 1:1.3 "normal viewing distance". – nwsp Dec 11 '14 at 13:55
  • Right, but you are still assuming a static image not a moving one. Your brain blends frames. Think resolution gains from stacking lower resolution shots and doing analysis. Your brain does this on the fly with video. It is a lot more obvious if you use SD video and look at the details of something on a still frame vs moving, though part of that is also the interlacing. – AJ Henderson Dec 11 '14 at 13:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.