How many different interchangeable lenses do you have?

Started 2 months ago | Polls thread
fferreres Veteran Member • Posts: 6,894
Re: OT: DoF measurements

ProfHankD wrote:

fferreres wrote:

Hank, is there a way to do a DOF test for lenses? I am a bit obsessed into the different transitions in the OOF PSF very close to the focus plane.

I think you know that it's my fault that OOF PSF is a thing... normally, PSFs are measured in focus. However, you're right that wacky things happen in the transition. OOF PSF is largely determined by simple geometry, whereas (in-focus) PSF is generally dominated by diffraction effects and aberrations. In gross cases, the OOF PSF is still very significant in DoF -- for example, the doughnut OOF PSF of a mirror lens cause such lenses to have a sharper transition to out-of-focus and less DoF.

The standard way to measure DoF is by photographing a ruler, but reading DoF from the image isn't an automated process. I'll add this to my list of things to think about: is there a simple test that can accurately and automatically generate a DoF scale?

Oh my, it;s a boring century when all the ideas are about using computer vision? CV has become very good at recognizing text (or features) in images, and many algorithms may are made a web service or can be downloaded as a trained model. Usually, they can be made to answer what is their confidence level per detected object.

If instead of a ruler we have a soup alphabet, or some abstract even smaller features (eg, cross, minus, plus, etc), maybe even with features of random sizes to test recognition by different frequencies, a software could be made to automate one synthetic perceptual reading without any human input and automatically.

Since the information is shot with the target at an angle, provided it's fully enclosed in the picture, the algorithm could infer the exact angle, and if the ruler is a fixed target size, I'd say A4 as the most common paper size in existence, then it could automatically extract the shooting distance, focal length (at least a range), angle, estimate the actual z-depth in the target and do all the calculations automatically, then straighten the target (like mobile scanners do) undoing perspective, then run the OCR or Computer Vision model, with information about the position of the letters, with K-means could infer the center cluster (sharpest area) or maybe this is much simpler - this case is so that the test could be done without even having to properly focus perfectly on the middle of the sheet.

Then how to generate a chart/scale based on readings of objects and the confidence level. I think it needs to use chracters where some confidence is actually good enough for something like small detail of a camera. Eg, digital numbers could be bad as they are so simple it could have high confidence and be wrong due to how things may align to the sensor)

The principle would be that if it cannot be guessed with any certainty, then the information is lost to blur. A benefit of the test is that I'd be explainable, and be able to be corroborated by humans. If I show you three pictures of a ruler with marked numbers, we'd likely agreed. And if disagree, it's because maybe we thing we can read it, but its just because we ordered the numbers in the ruler and know already what is the next number. That's why I think of a randomized alphabet soup.

If feel a reason to like a lot of lenses is the different rendering of this aspect, of which the very far away parts just an anecdote. I say this, because I think that even looking at ALL the measures/tests we have, like vignetting, MTF, curvture, etc. the character of a lens is completely missed by all this tests. I recall your suggestion to shoot the PSP with a tiny light, maybe shoot different parts of the frame with it.

The OOF PSF concept is why I started buying too many lenses in the first place. In fact, one of my earliest talks on this is still around as a video of a talk I gave at Microsoft in 2011: The Benefits Of Being Out Of Focus: Making the Most of Lens PSF ....

I think it's the most fascinating aspect. While many care only about what's in focus (obsession with the high cotnrast high MTF readings that LensRentals never gets tired of testing) another faction is obsessed by everything but it. But that the mechanism that produces all that's not in focus is the OOF PSF or the shape of light in between completely defocused and perfectly focused.

Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow