Patent: Canon Hexa Pixel CMOS AF

[Canon Rumors Guy]

A patent showing advancements in CMOS based autofocus from Canon has appeared. This patent seems to address an advanced pixel based contrast and phase AF, and even using the two together for focus adjustment. This could also be an advancement to Dual Pixel AF that Canon currently uses in various cameras across their lineup.</p>
<p>Patent Publication No. 2015-121778 (Google Translated)</p>
<ul>
<li>Published 2015.7.2</li>
<li>Filing date 2013.11.21</li>
</ul>
<p>Canon patents</p>
<ul>
<li>Image plane phase difference AF using 16 sub-pixels</li>
<li>Light field contrast AF that uses a</li>
<li>It does focus adjustment (Microadjustment) from the difference between the image plane phase difference AF and contrast AF</li>
</ul>

 

[siegsAR]

Hexa what? Can somebody explain that to me? ;D

Light field focusing, seems interesting.

 

[neuroanatomist]

If driven by 16 subpixels, should be hexadecapixel (hexapixel would be 6 subpixels).

 

[GMCPhotographics]

Wouldn't it be cool if you could use the live view phase detection AF to calibrate the regular AF system for each lens?

 

[ScottyP]

Maybe they think "Hexa Pixel" is a more catchy name.

 

[Stu_bert]

Wouldn't it be cool if you could use the live view phase detection AF to calibrate the regular AF system for each lens?

I thought it was contrast AF which doesn't need AFMA and works in live view, Phase being the normal AF?

Either way, yes please, in the 1Dx MK II. It would be awesome if it could do phase detection initially for speed and flip to contrast detection to lock in on the eyes

 

[neuroanatomist]

Maybe they think "Hexa Pixel" is a more catchy name.

I think petapixel already won that race.

 

[Etienne]

Whatever it is, if it can improve AF performance in every area, then Bring It On. I am hoping for a grand-slam home run in the 5DIV

 

[Mt Spokane Photography]

I see two things in the patent. Which is it? They seem to be totally different goals made possible by the same physical sensor design.

1. A calibration mode that uses contrast detect to memorize a correction for phase detect autofocusing, and applies the correction as in AFMA. At the time focusing is performed a quick check of focusing accuracy is done.

"According to this example, the calibration mode (adjustment mode) for adjusting a focusing position can be performed. A user can use a highly precise micro adjustment value in comfort with constituting the focusing position used as the maximum of a contrast value for high degree of accuracy and a short time, so that acquisition is possible. Thus, according to this example, it becomes possible to perform highly precise focal positioning at a high speed more. "

2. A Lytro light field like process that allows focusing after the image is captured using a computer.

"The refocusing technology of changing the focusing position of a captured image after photographing is known. Refocusing technology is carried out by photoing simultaneously a plurality of parallax images which carried out pupil division of the exit pupil of a taking lens to a plurality of regions, and responded to the divided pupil area. A Patent document 3 has disclosed the image sensor provided with one micro lens and a plurality of divided photoelectric conversion parts to one pixel. A plurality of parallax images which responded to a pupil area which is mutually different using each signal which carried out light reception by a plurality of divided photoelectric conversion parts are generable. A Patent document 4 has disclosed the imaging device which adds all the signals which carried out light reception by a plurality of divided photoelectric conversion parts, and generates a captured image.

A plurality of photoed parallax images are equivalent to the Light Field (LF) data which is the spatial distribution and the angular distribution information on light intensity. The nonpatent literature 1 has disclosed the refocusing technology of changing the focusing position of a captured image after photographing by synthesizing the image in a different virtual image formation face from an imaging surface using LF data."


AH This makes it all obvious.

 

[mclaren777]

1. A calibration mode that uses contrast detect to memorize a correction for phase detect autofocusing, and applies the correction as in AFMA. At the time focusing is performed a quick check of focusing accuracy is done.

I've been eagerly waiting for Canon to make use of this patent.

If the 5D4 uses this technology, I will surely buy two of them.

 

[COBRASoft]

Would this help with the focus issues people report for Sigma stuff?

 

[Don Haines]

Maybe they think "Hexa Pixel" is a more catchy name.

I think petapixel already won that race.

I thought it stood for pixels of our pets......

 

[Don Haines]

Wouldn't it be cool if you could use the live view phase detection AF to calibrate the regular AF system for each lens?

The article here seems to indicate it does, no?
"It does focus adjustment (Microadjustment) from the difference between the image plane phase difference AF and contrast AF"

That would be handy!

I predicted this when dual pixel tech was first released...... It just had to be coming.... It is such a logical use of the technology that it is inescapable..... Plus, it allows more AFMA points than just the ends of the Xoom range to be calculated and allows AFMA at different distances as well.....

 

[Mt Spokane Photography]

Would this help with the focus issues people report for Sigma stuff?

All third party lenses pretend to be a Canon lens and send a code for the Canon lens to the camera. Ideally, their characteristics would match the Canon lens they are emulating, but we know they do not.

Since each Canon camera has a internal data table that has information about focusing of each Canon lens, it already creates a AF issue for third party lenses, since that table may be different for different camera sensors, and make it a compromise for switching a lens from crop to FF, for example.

This could create yet a different internal table with expected peak MTF values for each lens, and that possibly would not play well with third party lenses (my guess). I did not quote that part, and I'm not going to dig it out again, but I recall it being in the patent.

 

[GMCPhotographics]

Would this help with the focus issues people report for Sigma stuff?

I think that's generally improved by applying sandpaper to the objective lens and then hitting the lens body with a large hammer

 

[Mt Spokane Photography]

...
This could create yet a different internal table with expected peak MTF values for each lens, and that possibly would not play well with third party lenses (my guess). I did not quote that part, and I'm not going to dig it out again, but I recall it being in the patent.

But given the variance between the same model lenses in resolution, I don't know how well that would work without the range being rather large. Or a huge improvement in Canon's manufacturing processes for lenses.

Actually, of around 100 Canon lenses I've done a AFMA on, only 2 or 3 were more than 10, one was 14 and one was 17. I'm sure there are bad ones, but I have a lot of data points and see few issues.

The patent has added a calibrate mode to a camera, it uses phase detect to determine the focus mismatch, saves the correction in memory. Basically, its in-camera AFMA. As to whether it can calibrate for various distances, or various focal lengths for zooms, its not mentioned, thats likely a firmware thing.

 

[RobPan]

Maybe they think "Hexa Pixel" is a more catchy name.

I think petapixel already won that race.

As in pétomane? Yotta!

~~~~