On human visual acuity

In various areas of photographic shot and image delivery planning we consider the visual acuity of the human observer.

For example, in doping depth of field planning, under what I call "outlook A", we seek to have all significant scene features confined to a range of distances such that the blurring they suffer due to not being at the distance at which the camera is focused will not be be "significant" to the human viewer, in the image viewing context we have in mind.

Simplistically, we may define that criterion of "blurring not significant to the viewer" as when the blur figure from misfocus is of comparable dimension to the blur figure that characterizes human visual acuity.

If we seek to work out how this depth of field analysis should be done, we must then determine what that metric of human visual acuity is.

We quickly find it said that the acuity of the human eye "with normal vision" is one minute of arc (1/60°). But what exactly does that mean?

We soon are ensnared in the familiar 2:1 ambiguity that infests the entire area of "resolution" (lines vs. line pairs, etc.)

Often we hear it said that the human eye with normal vision can " resolve two lines separated by one minute of arc". What does that mean? It is normally taken to mean that if we have two black lines on a white background, and the center-to-center distance between the lines as viewed is one minute of arc, then the "normal vision" eye can just barely recognize them as distinct.

Is that a correct interpretation? Probably not. Is that an accurate statement? I doubt it.

If we follow the trail of the definition of normal vision as tested with the classical Snellen eye chart, we find this matter very clearly defined in the literature (although I will express it here in my own words, which most directly relate to our outlook):

The characters (optotypes) on the modern Snellen chart are "Sloan letters" (after the typeface designer), and (except for the curved outlines and diagonal strokes) we would describe them as being defined on a 5 pixel high grid.



Normal vision is defined as the subject being able to correctly identify five out of six Snellen optotypes when their total angular height as seen by the subject is five minutes of arc; that is, when the vertical "pixel pitch", or "pixel size", of the optotypes is one minute of arc.

This "normal" visual acuity is said to have a "Snellen fraction" of 1.0; this is usually stated when working in the context of "inch" units as "20/20; when working in the context of SI units, as "6/6".

A person that can at best distinguish, at a distance of 20 feet/6 m, the row of optotypes with the green bar below is said to have 20/30 (6/9) acuity.

The literal meaning of the notation is that this subject can distinguish, at a distance of 20 feet (6 m), the optotypes a person with "normal" vision can distinguish at a distance of 30 feet (9 m).

The implication (in our terms) of an acuity stated as "20/30" is that the subject can recognize optotypes that vertically subtend 7.5 minutes of arc.

Best regards,

Doug

Now that we know how to deal with visual acuity, it would be interesting to compare "outlook A" and "outlook B" depth of field reckoning for the case of on-screen viewing,

I will use my situation as an example. My screen is about 1700 px × 1000 px, about 17" x 10" - thus the screen resolution is essentially 100 px/in.

Suppose the entire camera image is presented to my image viewer, which downsizes it to 1500 px × 1000 px, so it will nicely almost fill my screen for viewing (its dimensions will be 15 in. × 10 in.). If the camera output were 3900 px × 2600 px, that would be a 2.6:1 downsizing.

I always view my screen from a distance of 24 inches (I have no choice - that's how I designed my eyeglasses.)

At that distance, one minute of arc (the metric of typical human visual acuity) is 0.007 in. The pixel pitch is about 0.01 in. So we have a pretty good match between screen resolution and my acuity.

Now suppose that, predicated on this use to be made of the image, we do some depth-of-field planning for a shot.

Assume we have a 36 mm × 24 mm sensor.

Because the COCDL is defined at the focal plane, we must note that our image is magnified from the optical image on the sensor by a factor of about 10.6.

Under outlook A (based on visual acuity), I might use a COCDL of 0.007 in. at the on-screen image, or 0.017 mm at the focal plane.

Under outlook B (based on not compromising the potential viewed image resolution), I might use a COCDL of 0.01 in. at the on-screen image, or 0.023 mm at the focal plane.

Not a very big difference.

So in this case, the way I would do it, outlook A and outlook B would lead to broadly same shot planning from a depth of field basis.

Some however would feel that, under outlook B, we should us a COCDL that is the same as the camera sensor pixel pitch. In our example, that is 0.009 mm (at the focal plane, of course).

That means that the blur circle for an object point at what we would decide was the permissible limit of object distances would have a diameter of 0.009 mm on the sensor.

Now, on-screen, that diameter would be 0.0038 in - about 1/3 of a pixel.

So the blurring that we "allowed" on the sensor (by, for example, our choice of aperture, based on the depth of field reckoning) would be represented by a blur circle that, on-screen, was only 1/3 pixel in size. Not even close to having any effect on the viewed image.

So we have "left a lot of something on the table".

But, by God, at the camera output image, we have not allowed blurring from imperfect focus to compromise the resolution of our camera!

Best regards,

Doug

Doug Kerr said:

Under outlook A (based on visual acuity), I might use a COCDL of 0.007 in. at the on-screen image, or 0.017 mm at the focal plane.

Under outlook B (based on not compromising the potential viewed image resolution), I might use a COCDL of 0.01 in. at the on-screen image, or 0.023 mm at the focal plane.

Not a very big difference.

Click to expand...

Hi Doug,

Indeed, not a big difference, and only different because the display adds a physical limitation to resolution on top of what we accept as an acceptable amount of blur. The same exercise for printed output, where the output PPI jumps from 100 PPI to (upto) 720 PPI, would not show such a further limitation. Instead it would open up several ways of enhancing the image, including its 'perceived sharpness' (but optimizing output quality is something for another thread).

So in this case, the way I would do it, outlook A and outlook B would lead to broadly same shot planning from a depth of field basis.

Click to expand...

Yes, as my planning tool also shows, when the goal is to produce screen size output:

Some however would feel that, under outlook B, we should us a COCDL that is the same as the camera sensor pixel pitch. In our example, that is 0.009 mm (at the focal plane, of course).

Click to expand...

I have difficulty understanding why one would feel that. Sensel pitch has little to do with output resolution, but it has everything to do with how densely the projected image is sampled. Sensel pitch, or rather sampling density, does determine the theoretically highest possible resolution one can achieve. Of course there also other factors play a role, like the combined system MTF and deconvolution sharpening, and viewing conditions.

My model will also tell (when the output size is not manually changed) that that could produce a maximum output size of 27.23 x 18.15 inch (= 3900 × 2600 px @ 143.24 PPI), and when viewed from the same distance, its highest resolution (in the plane of focus) would still have the same resolution to match our visual acuity. But that's a different goal than the screen display goal from the beginning of your post..

Therefore, because we are now looking at a much larger image, we are also looking at proportionally larger blur diameters for subject matter that's not in the plane of focus. If we were to tolerate no larger blur diameter than in the smaller image for subjects a given distance in front, or behind, of our plane of best focus, we then need to adopt the use of a different COCDL to achieve that.

My model also automatically provides that, and allows to determine the required aperture, or focus distance, or focal length, or focus stacking, to achieve that. However, those changes will also change other aspects of how our image will look, like infinity blur, or perspective, or field of view.

The creative photographer is still in control, by choosing his/her weapons carefully.

Cheers,
Bart

Hi, Bart,

All well said.

Thanks.

Best regards,

Doug