Focus/recompose on sensor

[giordano]

Advertisement (gone after registration)

Imagine this with your EVF autofocus R-replacement:

1. Aim the "rangefinder spot" at the object you want in focus.
   
2. Press the button or turn the ring to get it sharp.
   
3. Reframe to get the composition right.
   As you do so, the camera adjusts the lens so your "target" retains maximum sharpness.
   

Easily done with on-sensor focus detection and today's processing power. Much more certain than multi-point AF, and much less fiddly than pressing cursor buttons to move a focus point round the screen.

 

Any takers?

[stunsworth]

How does the camera know where the target is in the frame once you've moved the camera? The photographer knows because he/she knows how much they've moved the camera and can see the subject in the frame. For the camera to do the same it would need to know how much the angle through which the camera had been moved horizontally and vertically.

[mjh]

That would certainly be feasible with todays technology. While you wouldn’t be able to guess camera movement from the live view data, the gyro sensors required for any kind of image stabilizer would provide data about angular movement in two dimensions (pan and tilt).

[stunsworth]

But accurate to be able accurately place a focus point - which could be something quite close and small with the background a long way away?

 

I'm sure there's military speced hardware and software that can do this, I'm just not sure about at the consumer level.

[mjh]

But accurate to be able accurately place a focus point - which could be something quite close and small with the background a long way away?

The accuracy required is much lower than what an image stabilizer requires, so even the kind of gyro sensors used in consumer level cameras would probably suffice. There’s just one issue: To assist in focus-and-recompose, angular movement would have to tracked over a couple of seconds rather than just a fraction of a second (as required by an image stabilizer), so small errors might accumulate over time. But since the precision required is much lower, this shouldn’t be an issue.

[giordano]

How does the camera know where the target is in the frame once you've moved the camera? The photographer knows because he/she knows how much they've moved the camera and can see the subject in the frame. For the camera to do the same it would need to know how much the angle through which the camera had been moved horizontally and vertically.

 

A few possibilities:

• Gyros (if they're good enough to drive image stabilisation they're more than good enough to move a focus point) which work out the angle through which you've moved the camera, work out the cosine, get the focus distance from the lens and calculate the adjustment that way.
  
• Gyros again, but this time used to track the focus point so the AF system can refocus after you've reframed.
  
• Computer power to track the movement of the image across the sensor, following the point you focused on.
  

The second and third methods have the advantage that they would compensate for any small inadvertent movement towards or away from the subject as you reframed - which is a big problem when using focus/recompose when DOF is very small.