720p windowed Resolution option for 1:1 focus & easy adjustment of back focus???

[PaulClements]

With regards to the "Day 1 of RED tests... unsharp?" thread I'd like to go back to the 1:1 issue. Whilst I hate banging a drum it's something that I strongly believe this camera can do easily and should be doing at present.

With the introduction of Anamorphic modes and the latest 2.40:1 mode it's obvious that with a bit of playing about Red can get the camera to shoot using certain pixels of the sensor only.

With this in mind cropping 1280 x 720 (Or 720p) from the center of the sensor would give everyone a live feed of 1 to 1 pixels to their monitoring.

The advantages of this would be:

• Hitting optimal focus practically every time
• Easy adjustment of back focus in controlled environments
• The ability to swap out mounts in the field with great accuracy
• The ability to check whether a lens has it's own focal issues
• The option to adjust on the fly to suit a lens that has such issues and there is no viable alternative on set
• A simple and reliable means of visualizing back focus adjustment so that cameras leave RED HQ exactly set from one to another
• An equality throughout Red footage

The only reason I can see that Red hasn't done this at present is because the camera can appear slightly soft at 1:1 due to the OLPF and lack of in camera sharpening and is therefore reticent to include this feature in case those new to the Red One judge it as being poor when they want to shoot 720p. Or perhaps add weight to the argument that the camera is not truly shooting 4K. But everyone who has used the camera quickly knows about the limitations and is happy to work with the footage at it's optimal sharpness.

Call the resolution type TEST or 1:1 FOCUS or something similar and include an option of a codec error if people try to record using it, just as it does when zoomed in currently.

Please don't leave this out as a feature to be included in Epic. It'd be a huge shame.

Cheers
Paul

 

[Chris Kenny]

I suspect the reason Red hasn't offered 1:1 display is because it requires a real debayer, whereas anything below 2K can be done by simply treating each block of four photosites as a single RGB pixel, which is much simpler and requires much less processing.

 

[Sanjin Jukic]

Thanks Paul.

A very good suggestion for RED ONE.

That was missing on RED from a day ONE.

 

[PaulClements]

I suspect the reason Red hasn't offered 1:1 display is because it requires a real debayer, whereas anything below 2K can be done by simply treating each block of four photosites as a single RGB pixel, which is much simpler and requires much less processing.

Perhaps, but I don't see why it would require a real debayer given what I'm suggesting uses the same process that gives you 2K or 3K or anything other than 4K. I'm not talking about scaling down to 720p from the full sensor, I'm talking about cropping 720p from the center.

As you say a single pixel is made up of the four photosites. The camera then has to create a 720p version of the 4K, 3K, 2K file or whatever. By contrast using the same four photosites as a pixel and taking 1280 x 720 as a crop from the center of the sensor and then outputting a 720p signal to the monitor should actually require less processing power on the part of the camera because it's not having to scale each frame down as much.

I'd very much like to hear from anyone from Red about this. Perhaps I'm over simplifying things and a definitive answer would be welcomed.

Paul

 

[Stuart English]

1:1 scale for focus

1:1 scale for focus

I'd very much like to hear from anyone from Red about this. Perhaps I'm over simplifying things and a definitive answer would be welcomed.

Paul

Paul, we do hear you.

Its not a simple task, and that is the only reason that this has not been done.

 

[Dan Hudgins]

Monochrome 1:1.

Monochrome 1:1.

I suspect the reason Red hasn't offered 1:1 display is because it requires a real debayer, whereas anything below 2K can be done by simply treating each block of four photosites as a single RGB pixel, which is much simpler and requires much less processing.

There would be no need for a de-Bayer since 1:1 or 1:2 (one sensor to four monitor pixels) would be used for setting the focus and OLPF to divide the light to four pixels.

A monochrome 1:1 would also let you see the RED/BLUE color balance so that you can put filters on the camera to get RED/BLUE balance to get max dynamic range, you just put a gray card in front of the camera out of focus and the color balance is right when the Bayer checker board is no longer visable.

[BTW: I need some sample 4K and 2K 16:9 TIFF files made by build 15 and 16 RED software to help de-bug my 48bit conversion, any help in this area would be great...]

 

[Chris Kenny]

As you say a single pixel is made up of the four photosites.

No, not at 1:1. That's the point. You can use that shortcut if you want an image with a pixel count that's a quarter of your photosite count (e.g. 2K from 4K), but at 1:1 your pixel count has to be the same as your photosite count.

Taking 1280x720 photosites and producing a 1280x720 RGB image requires a debayer. A debayer is inherently required to view images from a bayer-pattern sensor at 1:1. Unless, of course, you don't mind looking at an image with a visible bayer pattern on it, as Dan suggests. How useful that would be for focusing is hard to say; I'm not sure anyone has ever tried it. The sharp boundaries created by the checkerboard pattern might make it hard to judge how sharp your actual subject is.

 

[Dan Hudgins]

With filter no bayer pattern

With filter no bayer pattern

No, not at 1:1. That's the point. You can use that shortcut if you want an image with a pixel count that's a quarter of your photosite count (e.g. 2K from 4K), but at 1:1 your pixel count has to be the same as your photosite count.

Taking 1280x720 photosites and producing a 1280x720 RGB image requires a debayer. A debayer is inherently required to view images from a bayer-pattern sensor at 1:1. Unless, of course, you don't mind looking at an image with a visible bayer pattern on it, as Dan suggests. How useful that would be for focusing is hard to say; I'm not sure anyone has ever tried it. The sharp boundaries created by the checkerboard pattern might make it hard to judge how sharp your actual subject is.

If you put a filter over the camera to white balance, and you use a monochrome test chars for focus, as most are, there is no bayer pattern with equal gain on the sensors, neutral color balance. So the bayer pattern is not an issue, just use a filter or daylight and a monochrome test chart.

 

[PaulClements]

Monochrome 1:1 would still be very useful. I presume each photosite would need to be switched from red, green or blue channels to an equal grayscale channel and this is where the trouble lies perhaps?

Thanks for chiming in Stuart... Guess I'll have to stop banging that drum

Paul

 

[Mike Prevette]

Yes I fully support a Monochrome 1:1 In fact I'd prefer it to a color one because you would always know instantly that your in the punched in mode.

 

[David Groundwater]

Yes I fully support a Monochrome 1:1 In fact I'd prefer it to a color one because you would always know instantly that your in the punched in mode.

good thinking, mike.

since having some serious problems getting sharps with a set of master primes (discussed in another thread), i've begun to think that 1:1 zoom is the single most important feature missing from the red.

cheers,

david

 

[Dan Hudgins]

Monochrome is RAW sensor output to monitor.

Monochrome is RAW sensor output to monitor.

Monochrome 1:1 would still be very useful. I presume each photosite would need to be switched from red, green or blue channels to an equal grayscale channel and this is where the trouble lies perhaps?

Thanks for chiming in Stuart... Guess I'll have to stop banging that drum

Paul

Monochrome 1:1 (zoom mode) on the monitor is just the RAW sensor output applied to all three video inputs on the monitor.

The chip in digital camera is just a monochrome CCD with the Bayer filter over it. If you use the right color filter over the lens and have the chip gains set equal and point the camera at a monochrome photograph you would see an image ABOUT 4x (4x by area, 2x linear) as sharp as in color since the OLPF is not needed (for the 1:1 zoom focus mode, but would be there to have the backfocus effects included). With the OLPF in place the image will be fuzzed up since the OLPF splits the image rays into 4 with the 4 rays going to four pixels, if it is adjusted properly.

Since the OLPF is not fully effective some driect ray gets through to the "center pixel" with fuzz going to the ones around it, so in Monochrome mode 1:1 display will show some detail in a monochrome image at full sensor resolution.

The Monochrome 1:1 (zoom mode) is also very useful for balancing the light to get maximum sensor dynamic range, you put a gray card in front of the camera, punch up 1:1 monochrome (zoom) mode (or 1:2 or 1:4 so you can see the pixels bigger) and change filters until all of the bayer checker board goes away and you get solid gray.

One reason why they may not want this feature is that the sensors have many "defects" that need to be "hidden" with a bad pixel map. Since the monochrome 1:1 mode would show the sensor defects you could see which cameras have better sensors?

The sensor probably does not have RGB channels, there is just monochrome output 12 bits from the sensor (unless it has for outputs, then it has four monochrome outputs). At any rate the sensor probably already like all other CMOS sensors has just monochrome output already, the color is added later in the de-Bayer.

Anything that shows single pixels would be a help in focus, you could have a switch to select the top, middle, and bottom 8 of 12 sensor bits to adjust for the subject brightness.

Since the LCD monotor does not have RGB for all pixels, you will get color Moire on the monitor when the white balance is off from the checker board. The way to avoid that is to make the zoom larger and use a 1:2 or 1:4 zoom mode with one sensor pixel being 4 or 16 monitor pixels.

 

[Uli Plank]

… point the camera at a monochrome photograph you would see an image about 4x as sharp as in color since the OLPF is not needed. With the OLPF in place the image will be fuzzed up since the OLPF splits the image rays into 4 with the 4 rays going to four pixels, if it is adjusted properly.

Excuse me, but this shouldn't stand uncorrected. The OLPF has nothing to do with the Bayer pattern, and it's needed for monochrome as well. It's needed to avoid running into the Nyquist frequency, which would introduce massive aliasing effects.
The Bayer pattern filter is another filter on top of single pixels filtering the individual color information for them.
The OLPF could be laid out a bit less conservative for monochrome – assuming you want to shoot black and white only – since you don't need to take chroma moiré into consideration.

 

[Dan Hudgins]

OLPF and the hue of small objects image through Bayer filter

OLPF and the hue of small objects image through Bayer filter

Excuse me, but this shouldn't stand uncorrected. The OLPF has nothing to do with the Bayer pattern,

The Bayer pattern is in a square of 4 pixels, the OLPF matches that pattern in one way or another, that is what they have to do with each other.

There are different kinds of filters that can act as a OLPF filter, but for use with a Bayer sensor one kind divides the light into four beams, one for the red, one for the blue, and two for the greens.

You would not want to use that on a monochrome sensor since it would reduce the resolution, or at least you would want it to be closer to blur the image 1/2 pixel rather than a whole pixel.

What the monochrome being talked about here is a 1:1, 1:2, 1:4 pixel zoom preview, which would work better without any OLPF except for the need to be aware of its effect, any alasing is of no consequence since the monochrome being talked about is just for the monitor in zoom mode, and does not get used, the sensor data is compressed for recording and de-Bayered in the normal way.

The OLPF is needed for the Bayer pattern to reduce the color moire otherwise small points of white light would just illuminate one of the pixels in a four Bayer pixel, think of a small white spot like a star in focus moving over the sensor, if the star image is on the red filtered pixel you get a red signal after de-Bayer, then it moves to the blue pixel and you get a blue signal after de-Bayer, the OLPF need to blur the star image to cover four or at least three pixels so that you get a white signal rather than a colored one for points of light. This is an awful artifact that can happen from Specular reflections on various subjects. If you low pass the chroma in the de-Bayer then you cannot reproduce small colored objects with the right hue, the OLPF is needed to resolve the hue of small objects when using a Bayer filter.

Here is some information on a four point OLPF,

http://www.epsontoyocom.co.jp/english/product/Optical/set08/olpf_4/index.html

http://www.epson.co.jp/e/newsroom/2007/news_20070906.htm

I hope this clears up the misunderstanding. The point I was making was that the OLPF makes getting the focus right a little harder, but you need to live with that since it is going to be there since it is needed for the Bayer filter to reduce color moire. For finding the right color balance filter it is better NOT to do the de-Bayer and to see the checker board as a monochrome display enlarged, that way you can know when you have filtered the red/blue balance, and so get the best dynamic range.

If you were shooting with a three color stripe sensor like the Panavision (tm) rather than a Bayer filter sensor, then you would optimize the OLPF for that pixel pattern, so they, the Bayer filter and the OLPF, are related and should be concidered two parts of a system that records color.

I did make the point that the OLPF does not block all of the light and so some gets through to allow for the de-Bayer and masking to give single pixel responce, but the more single responce you let through the more Aliasing and color Moire you get. Some people seem to have the idea that the de-Bayer resolution in RED and BLUE are equal to single pixel since you can mix in some of the luma from the green sensors, if you think about that what green part of the luma do you get from a red and black subject, none, so where does this detail you seem to think is there come from? On the Bayer sensor you in fact get 1/2 the resolution relative to a monochrome sensor on a red and black subject, or blue and black subject, than you would from a monochrome sensor, except for whatever luma you can get out of the green and blue sensors due to their filters not being pure. Since the OLPF needs to be heaver for a Bayer filter sensor than a monochrome sensor, the luma for red and black subjects on the blue and green sensors would be blured anyway even if you could extract it.

Anyway, all of this is off the point of how best to arrange the monochrome display of a zoom mode that lets you see the single pixels for use in focus and color balance adjustments.

 

[Uli Plank]

Thanks for the clarification.