Doug Kerr
Well-known member
Hi, Bart in particular,
I remain intrigued by the issue of depth of field calculations with regard to the situation where the (only) "product" of interest is downsized from the camera resolution.
Let's for the moment not concentrate on this premise:
Now, when we apply this outlook B to a case where the image being considered is the image directly from the camera, we seem to have determined, empirically, that this objective will be met when the near and far limits are calculated using the classical DoF equations with a value for the parameter COCDL that is 1.0 times the pixel pitch. I will proceed on that basis.
Case 1
Suppose that our camera sensor layout is such that it has 3000 pixels per picture height. Then we would use a COCDL of 1/3000 PH. Certain near and far limits would be indicated. We honor those in placing our objects, and find that in the image our objective has been met: the objects at the near and far limits do not have "significantly" less sharpness than the object at the focus distance.
That is, the additional blurring caused by imperfect focus for the objects at the field limits just barely came into significance against a "perfect focus" resolution implied by 3000 px/PH.
Case 2
Now consider the same same shot with a camera whose sensor layout has 1000 px/PH. Then, to pursue that same objective, we would use a COCDL of 1/1000 PH for DOF calculation. The near and far limits indicated would be farther from the focus distance than in the first case (or, to give the same indicated limits, we could presume the use of a larger aperture).
Suppose we in fact use that larger aperture, and have the subjects positioned just as in the first case.
Again, we should find that in the image our objective has been met: the objects at the near and far limits do not have "significantly" less sharpness than the object at the focus distance.
That is, the additional blurring caused by imperfect focus for the objects at the field limits just barely comes into significance against a "perfect focus" resolution implied by 1000 px/PH.
Case 3
Now in a third case we use the original camera, with 3000 px/PH, and then (outside the camera) downsample the image to 1000 px/PH.
It seems to me that the amount of optical blurring (in the camera) that, in case 2, resulted in blurring of the objects at the field limits that was just noticeable against a a background resolution implied by 1000 px/PH would here cause that same impact blurring of the objects at the field limits that was just noticeable against a a "perfect focus" resolution implied by 1000 px/PH)
Thus, to choose an aperture for this shot (assuming the same positioning of the principal objects as in cases 1 and 2) we would use the DoF equations with a COCDL of 1/1000 PH - a COCDL based on the "simplistic" resolution of the "product" image, not that of the camera sensor.
Bart, your extensive actual experience in these matters may indicate that such is not so, in which case there is some unsolved mystery to me (perhaps some gross flaw in my reasoning, or perhaps the subtleties of the reality of combining multiple MTFs differing between cases 2 and 3).
Best regards,
Doug
I remain intrigued by the issue of depth of field calculations with regard to the situation where the (only) "product" of interest is downsized from the camera resolution.
Let's for the moment not concentrate on this premise:
A. Our purpose in determining the near and far distances within which "important" subject features can be placed is that that they will suffer, from "imperfect" focus, blurring that will be noticeable to the human viewer assuming a certain combination of image size and viewing distance.
but rather we will concentrate on this premise:B. Our purpose in determining the near and far distances within which "important" subject features can be placed is that they will suffer, from "imperfect" focus, blurring that will not significantly degrade image sharpness compared to the sharpness for objects in perfect focus.
Now, when we apply this outlook B to a case where the image being considered is the image directly from the camera, we seem to have determined, empirically, that this objective will be met when the near and far limits are calculated using the classical DoF equations with a value for the parameter COCDL that is 1.0 times the pixel pitch. I will proceed on that basis.
Let me also assume that we are operation with an aperture such that diffraction blurring can be ignored.
Case 1
Suppose that our camera sensor layout is such that it has 3000 pixels per picture height. Then we would use a COCDL of 1/3000 PH. Certain near and far limits would be indicated. We honor those in placing our objects, and find that in the image our objective has been met: the objects at the near and far limits do not have "significantly" less sharpness than the object at the focus distance.
That is, the additional blurring caused by imperfect focus for the objects at the field limits just barely came into significance against a "perfect focus" resolution implied by 3000 px/PH.
Case 2
Now consider the same same shot with a camera whose sensor layout has 1000 px/PH. Then, to pursue that same objective, we would use a COCDL of 1/1000 PH for DOF calculation. The near and far limits indicated would be farther from the focus distance than in the first case (or, to give the same indicated limits, we could presume the use of a larger aperture).
Suppose we in fact use that larger aperture, and have the subjects positioned just as in the first case.
Again, we should find that in the image our objective has been met: the objects at the near and far limits do not have "significantly" less sharpness than the object at the focus distance.
That is, the additional blurring caused by imperfect focus for the objects at the field limits just barely comes into significance against a "perfect focus" resolution implied by 1000 px/PH.
Case 3
Now in a third case we use the original camera, with 3000 px/PH, and then (outside the camera) downsample the image to 1000 px/PH.
It seems to me that the amount of optical blurring (in the camera) that, in case 2, resulted in blurring of the objects at the field limits that was just noticeable against a a background resolution implied by 1000 px/PH would here cause that same impact blurring of the objects at the field limits that was just noticeable against a a "perfect focus" resolution implied by 1000 px/PH)
Thus, to choose an aperture for this shot (assuming the same positioning of the principal objects as in cases 1 and 2) we would use the DoF equations with a COCDL of 1/1000 PH - a COCDL based on the "simplistic" resolution of the "product" image, not that of the camera sensor.
Bart, your extensive actual experience in these matters may indicate that such is not so, in which case there is some unsolved mystery to me (perhaps some gross flaw in my reasoning, or perhaps the subtleties of the reality of combining multiple MTFs differing between cases 2 and 3).
Best regards,
Doug