15 to 20 stop range with two RED ONE (tm)

[Dan Hudgins]

It should be possable to get a longer scale for daylight or high contrast lighting by using two RED ONE (tm) cameras and a 10%-90% beam splitter in front of the cameras lenses.

The cameras would need matched lenses, and the camera rig would need to be solid and have adjustment to compensate for the 1/3 thickness of the beam splitter so the nodal point of both lenses correspond.

The dimmer camera could be cut more with ND to get a spread of 5 to 8 stops between the cameras. You would not want an 11 stop spread since you need to curve over the high (and low) clip point before the two images are FUSED.

Both frame sets would need to be preped with curves to avoid any sharp clipping before being fused, and the output set would be run through curves and color matrix to get a more film like rounded curve.

Both cameras would need to be in sync.

Since many RED ONE (tm) owners already have two cameras, a two camera rig could be sold as an accessory with a 10%-90% beam splitter used for extended range shooting, and a 50%-50% beam splitter used for 3D stereo shooting.

The images would need to be cropped a little and resized to make up for registration errors. The iris should be the same in both cameras unless you want a sharp+soft image for effect.

To extend this idea, with three cameras and two beam splitters you could shoot through RGB filters, and with 9 cameras you could shoot extended range with RGB filters, although you could need long lenses and a field lens with the prime lens in front, not with a ground glass, just relay optics in order to get wide angle shots.

 

[Andrew M.]

And why not to genlock two cameras aligned (+-) 5 pixels and merge in Premiere using HDR script?
HDR merging technique now can compensate for pixel shift using edge detection very well, back to one pixel accuracy.

Even better is to have shutter speed bracketing implemented on the RED and shoot 48fps.

 

[dewaldaukema]

It should be possable to get a longer scale for daylight or high contrast lighting by using two RED ONE (tm) cameras and a 10%-90% beam splitter in front of the cameras lenses.


20 stops would look something like this.

 

[Gunleik Groven]

That's exactly why I find it hard to understand HDR...

 

[dewaldaukema]

That's exactly why I find it hard to understand HDR...

This dragged back to ±9-10 stops...

 

[istvanttt]

It should be possable to get a longer scale for daylight or high contrast lighting by using two RED ONE (tm) cameras and a 10%-90% beam splitter in front of the cameras lenses.

The cameras would need matched lenses, and the camera rig would need to be solid and have adjustment to compensate for the 1/3 thickness of the beam splitter so the nodal point of both lenses correspond.

The dimmer camera could be cut more with ND to get a spread of 5 to 8 stops between the cameras. You would not want an 11 stop spread since you need to curve over the high (and low) clip point before the two images are FUSED.

Both frame sets would need to be preped with curves to avoid any sharp clipping before being fused, and the output set would be run through curves and color matrix to get a more film like rounded curve.

Both cameras would need to be in sync.

Since many RED ONE (tm) owners already have two cameras, a two camera rig could be sold as an accessory with a 10%-90% beam splitter used for extended range shooting, and a 50%-50% beam splitter used for 3D stereo shooting.

The images would need to be cropped a little and resized to make up for registration errors. The iris should be the same in both cameras unless you want a sharp+soft image for effect.

To extend this idea, with three cameras and two beam splitters you could shoot through RGB filters, and with 9 cameras you could shoot extended range with RGB filters, although you could need long lenses and a field lens with the prime lens in front, not with a ground glass, just relay optics in order to get wide angle shots.

Wouw never thought about that, excellent idea, thanks

 

[istvanttt]

It should be possable to get a longer scale for daylight or high contrast lighting by using two RED ONE (tm) cameras and a 10%-90% beam splitter in front of the cameras lenses.


20 stops would look something like this.

love it

 

[Michael Stanmore]

All these things need to be used in moderation... I like the idea of exploring HDR cinematography further to achieve a certain painterly look, but it's gotta be done with skill and subtlety... or it'll just look like terrible compositing...

 

[HDPioneers.com]

I've been dreaming about HDR cinematography for awhile. Biggest hurdle seems to be running a script to merge the images intelligently, and not have it take days rendering. If someone will build this Rig I'd gladly lend my camera for testing

 

[hunterrichards]

Been on my mind for years- thought about getting a patent but decided not to- here is the better way:

Build a new camera with 2 cmos sensors, and a beam splitter (in the place where a viewfinder prism would go on a film camera). Lay each sensor on the back of the block (like a 3-chip RGB block)- computer aligned to match pixels. Record in sync, offload to 2 separate pieces of media (one from the "highlight" sensor, one for the "shadow" sensor- use something like Red Alert "HDR" to process the footage (flip "highlight" sensor on the H axis, tone map and export dpx). and POW! You have at least 17+ stops with great mid-tones as well (The key to this is just nailing the beam splitter to have the proper light transmission values)

Please send me a bone if you use this method! I have diagrams as well.

-Hunter

 

[Dan Hudgins]

Dithering to hold expanded scale on display...

Dithering to hold expanded scale on display...

This dragged back to ±9-10 stops...

When you go from 20+ stops range down to the limited range of a display you need to dither the tone values of you will get maybe 3-5 binary steps in the original image lost in one binary step of the display.

This is also an issue with the film recorder even if the film recorder has a 16bit D2A you still need dithering to get the tones out to the film print stock.

You will notice the heavy tone banding in your JPG sample image in the sky.

It is not possable to get the range of tones needed in a compressed image for soft out-of-focus areas like fog and mist. Only uncompressed images that are dithered at high resolution can average in the viewers eye to give smooth tones that are like film-to-film printing. 65/70mm is better for fog since the dithered points are closer together and make for a finer pitch blending of the tone values in the viewers eye.

When you change the slope of the expanded range image to fit the display the contrast will need to be lower than what the eye sees, so you would use an S curve that would increase the contrast in the mid-tones and reduce the contrast in the highlights and shadows which is what film use does.

To avoid losing detail you would dither the images so that tones are made up of about 9 pixel groups in the 4K image, that will help some of the tone information get through the film recorder onto the print stock (yes I know that people make printing negatives, but you will get better images going directly to the print stock it is sharper and has better color with less grain.)

 

[Dan Hudgins]

Shutter speed bracketing?

Shutter speed bracketing?

Even better is to have shutter speed bracketing implemented on the RED and shoot 48fps.

I don't think you want to use shutter speed to bracket the exposure since you want the shutter open and close to be the same for the high and low expsoures. Otherwise you would get "picket-fencing" in the short exposure image.

Using the shutter would be better for the IR leak problem with ND filters, but you don't want motion artifacts like in "Gladiators" all the time... do you?

Also if the images are consecutive things that move will take on a bright edge effect so the two images will not "cover" each other and the brightness curves will not blend for all pixels when things move.

 

[dewaldaukema]

Original file 56MB, the one you are looking at is 199 KB...

 

[Dan Hudgins]

Daylight...

Daylight...

All these things need to be used in moderation... I like the idea of exploring HDR cinematography further to achieve a certain painterly look, but it's gotta be done with skill and subtlety... or it'll just look like terrible compositing...

When shooting in Daylight you would be taking the 20 stops into an S curve, so I would want to use it ALL the time to avoid having the highlights blocked up on bright reflections, and get a more soft shoulder like film. Hard lighting makes reflections on surfaces that are much higher than a "white card" reading.

When I do black and white film I use developers that auto compensate for high exposure to hold the highlights better, since the sensor has a straight "curve" the only thing you can do is to extend it way high, then bend it back down in post, otherwise your shadow detail is lost in the big binary steps at the bottom of the digital signal.

 

[Dan Hudgins]

Just making a point about compression...

Just making a point about compression...

Original file 56MB, the one you are looking at is 199 KB...

Anytime compression averages two pixels or more you are losing tone values.

 

[P Andersson]

My guess is this will be achieved soon in cameras using only one sensor. I think there will be some way of using the space of the pixels in the same way that the eye operates, some pixels that are smaller are used for color accuracy and some sensors that are larger are used for seeing into the dark.

Here is a link to a Kodak 2007 new way of thinking in sensor design. They propose to remove the colored filters from some of the pixels to increase sensitivity.
http://johncompton.pluggedin.kodak.com/default.asp?item=624876

there was a comment after the article about putting nd filters for the extreme highlights on some of the clear pixels to increase dynamic range - i wanted to see what that would look like and also thought of making one larger pixel in the center for taking in more light in the shadows

 

[Dan Hudgins]

Two sensors are better than one?

Two sensors are better than one?

Been on my mind for years- thought about getting a patent but decided not to- here is the better way:

Build a new camera with 2 cmos sensors, and a beam splitter (in the place where a viewfinder prism would go on a film camera). Lay each sensor on the back of the block (like a 3-chip RGB block)- computer aligned to match pixels. Record in sync, offload to 2 separate pieces of media (one from the "highlight" sensor, one for the "shadow" sensor- use something like Red Alert "HDR" to process the footage (flip "highlight" sensor on the H axis, tone map and export dpx). and POW! You have at least 17+ stops with great mid-tones as well (The key to this is just nailing the beam splitter to have the proper light transmission values)

Please send me a bone if you use this method! I have diagrams as well.

-Hunter

I had that idea some time ago, I posted an image of a R+GB sensor arangement to get better resolution that using the Bayer filter since the R sensor could be monochrome and require a less heavy OLPF.

You could use two Bayer sensors also for split brightness range use, one with heavy RGB filters for the bright image and one with weaker RGB filters for the dark image, that way you could get better color without loss of speed since the beam splitter would be 5%-95% or 10%-90% and such.

The problem with putting the beam splitter behind the lens is that it needs to be thin and that can cause problems with vibration since it acts like a drum skin.

It is better to put the beam splitters in front of the camera lenses, and to use long lenses on the cameras into a field lens, then you can use a prime lens of any angle of view without the problem of the "angle of total reflection" coming into play (which it could with a behind the lens beam splitter).

In the illustration the GB checker board is closer together than the Bayer filter so gives better sharpness for all colors, and the Red sensor would make the camera about twice as sharp as using the Bayer filter for saturated colors.

Some might say that you cannot see detail in saturated colors, that is a mistaken idea from the original NTSC tests that used about a 12 inch image viewed more than 6 feet away or so, how much color and tone information you can see depends on how big the image is and how far away you are, if you get closer or make the image bigger you can see more detail in the colors.

The same goes for 4K images, if you sit at the back of a small narrow theatre you cannot see pixels seperated by one pixel if the screen is only 8 feet wide, even if the projection lens can resolve them. But if the projection is uncompressed and dithered by noise you will be able to see image tones beyond the range of the digital steps in the file format and projector circuits since pixel groups will average in your eye, just the way film grains average in your eye to make extra brightness tones. Film also has overall brightness fluctuation from frame to frame because of the way film is made and processed, so it might be good to vary the brightness and gamma of digital images a little from frame to frame as well, but you cannot do that with compressed data sometimes since the encoder filters out small frame to frame and pixel to pixel brightness fluctuations that add to the perception of image tones taken from the subject.

Older Image-Orthicon cameras used four sensors and gave fine color, they had RGB plus monochrome. You could do that with a two chip arangement like this with a Bayer plus Monochrome. The reply to that is that you can get luma from the Bayer, but that is not all together true since the OLPF reduces the contrast of small details, if you had a monochrome sensor you would not need to peak the luma that is fuzzed up by the OLPF, and for the chroma you could make the OLPF heaver to reduce chroma moire.

You can use 4 sensors and three beem splitters but not with PL mount lenses unless you put the splitters in front of the lenses.

Anyway, I tink you could build a fixture to hold 9 RED ONE (tm) cameras to make a RGBx3 camera that would make better images and still be at a cost that is not too high for high end work. It would be better if the cameras could be fit with sensors that do not have the Bayer filter, I do not know if the RED company can get monochrome sensors for multi camera, stop motion filter wheel, and film scanner applications, but when they get the RAW port working also having a monochrome sensor option would be good also to get higher quality images.

 

[Dan Hudgins]

Olpf?

Olpf?

there was a comment after the article about putting nd filters for the extreme highlights on some of the clear pixels to increase dynamic range - i wanted to see what that would look like and also thought of making one larger pixel in the center for taking in more light in the shadows

The OLPF normally splits the light into four beams for the Bayer filter, to use a more complex array the light would need to be split many ways making the lens OLPF filter issues very complex and would also make for heavy moire in the brightness and chroma.

Having more range in the luma does not help the chroma noise in the shadows, what you would want is big RGB pixels for the shadow and smaller RGB pixels for the highlights. Chroma noise is worse in the shadows, and chroma cliping is worse in the highlights. So you need clean chroma.

Using a monochrome sensor and a Bayer sensor with a beam splitter lets you get luma detail and low pass the chroma seperately so that you can use OLPF that are not the same on both sensors to optimize the results to have more detail without peaking.

Anyway, that is getting away from the today, of just using the two RED ONE (tm) cameras sitting in your place on a fixture to get better results than using a single camera. Even if you use a 50%-50% beam spliter and give both cameras about the same exposure, when you FUSE the two images you get less compression artifact and cleaner images.

I added a Bayer drawing with large RGB for low light and ND+RGB smaller for highlights, N_BAYER.JPG... You would probably need to use this without a micro lens array, because the small sesors would be more sensitive to the angle of the light and the focal point of the micro lens array...

 

[michael zaletel]

Dan & Hunter:

This forum never ceases to amaze me. Thanks for sharing these ideas in such a public forum. Before reading this, I had never thought about this possibility. The amount I don't know yet continues to outpace what I do know.



-shooter

 

[cinepost35]

Could you get some overall combined max resolution increase/enhancement beyond HDR with the combined image for big wide vista shots if you wanted?
Like enhanced 4k for 4K output.