Doug Kerr
Well-known member
When we discuss such matters as the impact of blur due to imperfect focus, we often seem to judge the impact of blurring characterized by a blur figure of a certain diameter with respect to with the "baseline blurring" that is one way to look at the "resolution" of the camera itself.
We tend to think that if the diameter of the blur figure created by imperfect focus is less than the diameter of the blur figure implied by the finite resolution of the camera, then the blur caused by imperfect focus will have essentially no effect on the sharpness of the image.
The concept is that when two sources of blurring are cascaded, the greater one prevails, and the lesser one causes no further degradation.
But that is not so.
If the blurring implied by the camera's finite resolution follows a classical model called Gaussian blur, and the blurring caused by imperfect focus for objects at some distance also follows the Gaussian blur model (both fairly reasonable assumptions), then the diameter of the blur figure that characterizes the overall system response is the square root of the sum of the squares of the diameters of the two "contributing" blur figures.
Suppose our sensor has 2000 sensels per picture height. Then its resolution will be somewhat less than 1000 cycles per picture height - let's assume 750.
Then suppose we adopt 1/2000 of the picture height (the sensel pitch) as our COCDL. Then, for an object at one limit of the calculated depth of field, the blur figure from a point on the object will have a diameter of 1/2000 of the picture height.
Thus (I'll spare you the algebra) the resolution of the system for that object is 600 cycles per picture height - a 20% decline due to the imperfect misfocus at our "limit".
So can we really say, then, that our choice of a COCDL equal to the sensel pitch means that the blurring due to misfocus for objects at the limits of the calculated depth of field does not consequentially degrade the image sharpness for those objects?
Beats me.
Best regards,
Doug
We tend to think that if the diameter of the blur figure created by imperfect focus is less than the diameter of the blur figure implied by the finite resolution of the camera, then the blur caused by imperfect focus will have essentially no effect on the sharpness of the image.
The concept is that when two sources of blurring are cascaded, the greater one prevails, and the lesser one causes no further degradation.
But that is not so.
If the blurring implied by the camera's finite resolution follows a classical model called Gaussian blur, and the blurring caused by imperfect focus for objects at some distance also follows the Gaussian blur model (both fairly reasonable assumptions), then the diameter of the blur figure that characterizes the overall system response is the square root of the sum of the squares of the diameters of the two "contributing" blur figures.
Suppose our sensor has 2000 sensels per picture height. Then its resolution will be somewhat less than 1000 cycles per picture height - let's assume 750.
Then suppose we adopt 1/2000 of the picture height (the sensel pitch) as our COCDL. Then, for an object at one limit of the calculated depth of field, the blur figure from a point on the object will have a diameter of 1/2000 of the picture height.
Thus (I'll spare you the algebra) the resolution of the system for that object is 600 cycles per picture height - a 20% decline due to the imperfect misfocus at our "limit".
So can we really say, then, that our choice of a COCDL equal to the sensel pitch means that the blurring due to misfocus for objects at the limits of the calculated depth of field does not consequentially degrade the image sharpness for those objects?
Beats me.
Best regards,
Doug