Focus assist on DSMC2

[Cameron McKinlay]

Cameron,

+) The more complex the edge detection -> the more calculations per processed pixel.
-> As simple edge detection may involve checking the Greyscale (luma) value of a pixel with each of the surrounding 8 pixels
-> More advanced algorithms can attempt to dynamically adjust this in 'low contrast' areas (but with a higher CPU cost)

+) The higher the frame rate (of displayed edge detection image) -> the proportionally greater the number of pixels processed

Arbitrary Example (with approximate maths)

If 8 maths operations were required per pixel, on a 32 Megapixel image at 4 fps
32 Million x 8 SUBTRACT operations = 32,000,000 x 8 x 4 fps = 1 Billion Operations a second (1 Ghz of Maths operations)

If this was done in a CPU, it would already consume a lot of power.
It would also create considerable memory contention (with other critical processes).
With hardware assistance, this could be reduce, but other compromises make be used to reduce the overal burden on the system.

I can't offer insight as to what RED have decided is a priority for real-time processing, but do note that edge detection can eat into a systems resources (especially if adaptive contrast detection are desired).

AJ

That sounds really impressive and technical. What it sounds like is we should all be giving Cinemartin a round of applause for accomplishing something at a $99 price point when RED is unable to do it at any price point.

While the camera may be recording higher frame rates, the action is still happening in real time and the monitoring is still happening in real time.
Sorry AJ, I just don't buy it that a $30,000 cinema camera doesn't have the technological ability to do colored focus peaking when it already has that silly over-sharpening feature. Just make it a color, done.

[Antony Newman]

You may have a point.

My guess is that dedicated screens companies bought IP Blocks to perform Sobel (or other) edge detection in hardware (ie in an ASIC).
My guess would also be that RED have carefully constructed an implementation of Edge Detection in a 'programmable logic array' (FPGA).

FPGAs are much faster & efficient than CPU, but still consume a lot more electrical power (and are slower) than ASICs.
There is also a limited amount of space within a FPGA that may been to be dedicated to other (more critical logic).




Perhaps post IPP2 rollout, reclaimed FPGA space could be used to implement a more capable Edge detection algorithm or 1:1 magnify-loupe-while-recording.

AJ

[Brody_B]

The edge focus check currently is helpful, but most of the time it takes me out of the scene just seeing black and white edges, this to me is perfect for the AC to get spot on focus, but not at all for contextual framing of moving subjects for a DP, and yes when in darker scenes it just doesn't show up at all. Alternatively the regular focus check is the other option to use but it is quite hard to see when outside in the sun and just isn't very easy to see. A colored peaking version with and intensity adjustment like the one my cheap monitors and stills camera have would be pretty awesome Red. Not sure why it hasn't been mentioned or implemented by now, I have also seen tons of people ask about it.

[Mar van de Sande]

No Luca you arent getting old i'm 15 and i'm having trouble focusing on Epic-w

[Luca Orlandi]

Hahaha! Thanks, but is i'm 35 and honestly my eyes need a better focus peaking.

No Luca you arent getting old i'm 15 and i'm having trouble focusing on Epic-w