Luminous Landscape Forum
Equipment & Techniques => Medium Format / Film / Digital Backs – and Large Sensor Photography => Topic started by: FredBGG on April 15, 2013, 09:01:15 pm
-
Here's an idea. Hasselblad has their multi shot back to capture all RBD at every pixel.
How about doing this with the monochromatic back with a three filter changing mechanism.
This could also be an option for shooting black and white where you want to be able to choose different filtration
for different areas of the image.
I think it would be quite a nice accessory for many photographers.
Some lenses are designed with internal select able filters. Maybe build this into certain lenses as an automated feature.
It could be an interesting Phase One alternative to the Hasselblad and Sinar multishot backs.
There could even be implications for color balance using different exposure times for the three filters.
The end result would be the ultimate monochromatic back for action and still life as well as the ultimate IQ back for non moving subjects.
I wonder if it would be possible to insert a changing filter array into the tech cam MFD lenses by Rodenstock and Schneider
180mp image WITH NO MOIRE 8)
-
Once I can get my hands on an IQ260 Achromatic, I am going to do exactly that using a true red, green, and blue filter set in front of the lens and then placing each of those files into the respective channels in Photoshop and see what happens along with playing with some other methods for doing this.
Though, seeing your inclination to adapt and modify things, I'm sure you could find a way to hack or adapt some sort of filter wheel into the mix....http://www.telescope.com/Orion-Multiple-5-Filter-Wheel-and-Color-Filter-Set/p/20020.uts?keyword=filter might be easiest with an Hartblei Hcam or Alpa FPS...
-
Here is a Leaf filter wheel:
(http://thumbs2.ebaystatic.com/d/l225/m/m1Snw_NWFZKq2DlmJzExgJA.jpg)
-
In a still life or non-moving situation (think art copy) , this should be ideal for the best reproduction of color.
However, for ANYTHING moving (think branches and leaves in a slight breeze), mis-registration will rear its ugly head, really fast.
The next best thing would be color film in 6x7 or larger :)
And no comping after the shot
-Dan
-
This technique used hundred years ago in Russia:
http://www.boston.com/bigpicture/2010/08/russia_in_color_a_century_ago.html
-
Fred,
What you suggest is a good idea, and has a long pedigree in medium format digital - the Leaf DCB and Volare (either with the Leaf filter wheel or the Sinarcam1's built-in wheel), the Megavision T2 and T32, and so on.
I actually recently landed an old Leaf Volare kit, exactly like the one you pictured, to play around with. I am more interested in using it as a monochrome full-spectrum sensor however, or perhaps with an IDAS LPS filter.
And of course the vast majority of higher-end astronomical colour images are taken in this way. Every colour Hubble Space Telescope spectacle you've seen used some tricolour filter combination - quite often narrowband filters centred on particular emission lines of gaseous nebulae.
Hasselblad has their multi shot back to capture all RBD at every pixel.
....
180mp image WITH NO MOIRE 8)
But it's important to note that using a filter wheel with a monochrome sensor is not the same as using a multi-shot back, and will not give you 3x 60MP = 180MP image. There is no spatial sub-sampling. It remains a 60MP image - just a better one than what a colour sensor delivers, because there is no de-Bayering interpolation; no guesswork of high-frequency detail and of the full colour at each pixel. There is also better signal to noise, because 3 exposures were merged. Compared to the standard RGBG (2 green, 1 red, 1 blue) Bayer CFA block of 4 pixels, the green S/N is sqrt(2) (1.4x) better, while the red and blue S/N are each 2x better. One could merge 3 exposures with a Bayer sensor too [but why would one bother?]
Ray
-
I wonder what would be the workflow for profiling and balancing the color of a shot produced this way... depending on the spectral quality of the filters used, color could be all over the place.
-
Kolor-Picker: You would create a composite TIFF, create a profile from that TIFF using standard profiling software, and apply the profile in script/batch/automation form in Photoshop. If you found the color was significantly off you'd create presets in the raw processor for each of the three "color" captures to get the colors closer to start with so the profiling isn't doing too much heavy lifting in the final TIFF.
This is an abstractly interesting idea. I have done the same with gum bichromates. It would make for a fascinating art project or personal pursuit. However, it is not something I would suggest anyone pursue for practical/commercial applications.
There is a very good reason that Leaf used to make multi-shot backs and no longer does. There are still some theoretical advantages to a multishot system, but with higher resolution capture and significantly improved raw processing math they have become vanishingly small; a good 80mp single-shot back with great raw processing leaves VERY little to be desired image quality, color, and moire-resistance wise.
-
The good thing about a DIY color-filter wheel is that you can choose your own filter. Why limit yourself to 3 primaries, or visible light? Why not 10nm passbands from near-IR and into UV? Or two filters at opposite polarization?
Here are the passbands of a 13-band multispectral filter:
(http://www.lumiere-technology.com/images/Products/Multispectral.gif)
Here is a color wheel from a DLP projector:
(http://upload.wikimedia.org/wikipedia/commons/thumb/0/01/DLP_Colorwheel_IMG_5507.JPG/320px-DLP_Colorwheel_IMG_5507.JPG)
-h
-
There is a very good reason that Leaf used to make multi-shot backs and no longer does. There are still some theoretical advantages to a multishot system, but with higher resolution capture and significantly improved raw processing math they have become vanishingly small; a good 80mp single-shot back with great raw processing leaves VERY little to be desired image quality, color, and moire-resistance wise.
Hasselblad 200MS no sensor moire
IQ180 does produce sensor moire.
It's not theoretical.
-
Hasselblad 200MS no sensor moire
IQ180 does produce sensor moire.
It's not theoretical.
All cameras produce some moire/aliasing, especially those that have no OLPF. "Proper" sampling would demand pre-filters that are practically impossible to realize via optical means.
The main benefit of sensor movement/super resolution should be increased spatial bandwidth in the color (difference) channels. Shooting red/blue woven fabric?
-h
-
The main benefit of sensor movement/super resolution should be increased spatial bandwidth in the color (difference) channels. Shooting red/blue woven fabric?
Hi,
There are 2 resolution related benefits to certain multi-shot sensor practices. One is the increase in resolution, due to the denser sampling. Another benefit is the fact that a half sensel-pitch offset, with a relatively large aperture of the sensels, will produce an overlap with the 'initial' sampling positions. That overlap will reduce the modulation of signals near Nyquist, a sort of AA-filter effect.
Therefore the two effects enforce one another, higher sampling density will reduce the chance of creating aliasing (it requires even smaller detail and better lenses/focus to cause aliasing), and overlapping area samples will reduce modulation near Nyquist.
It's a related principle that is also exploited by Epson in their 'staggered' sensor alignment in scanners, which do not use tri-linear (R/G/B) sensors, but bi-linear per color (6 in total for R/G/B) filtering with half a sensel physical offset. According to their patent (http://www.google.com/patents/US6924840) it was mainly to increase resolution and speed. The staggered offset sensels work fine in a scanning device, but in a single shot full frame capturing device one would require a physical displacement of the entire sensor array by a fraction of the sensel pitch, which slows down the operation and raises the bar for consistent lighting and vibration reduction.
The aspect of increased color accuracy with a monochrome sensor and a color filter wheel only addresses the full sensel pitch quality of color and eliminates the need for demosiacing, which will eliminate false color artifacts caused by different sampling densities between Green, and Red/Blue filtered sensels. Resolution will only benefit modestly, but still significant enough, by maybe 10%, slightly more than the loss due to demosaicing because other filter band-pass overlaps can be used.
Cheers,
Bart
-
Once I can get my hands on an IQ260 Achromatic, I am going to do exactly that using a true red, green, and blue filter set in front of the lens and then placing each of those files into the respective channels in Photoshop and see what happens along with playing with some other methods for doing this.
Though, seeing your inclination to adapt and modify things, I'm sure you could find a way to hack or adapt some sort of filter wheel into the mix....http://www.telescope.com/Orion-Multiple-5-Filter-Wheel-and-Color-Filter-Set/p/20020.uts?keyword=filter might be easiest with an Hartblei Hcam or Alpa FPS...
I wonder if the chaps at Alpa would be interested in building a filter wheel into a 17mm adapter.
Now that would open up a whole range of intriguing opportunities...
-
Hi,
There are 2 resolution related benefits to certain multi-shot sensor practices. One is the increase in resolution, due to the denser sampling. Another benefit is the fact that a half sensel-pitch offset, with a relatively large aperture of the sensels, will produce an overlap with the 'initial' sampling positions. That overlap will reduce the modulation of signals near Nyquist, a sort of AA-filter effect.
...
I was commenting the Hasselblad method, which I thought only did the first (moving the sensor at integer sensel distances so as to cover a full color image using a Bayer color filter)?
Applications such as Photoacute does something more akin to the latter, except that the sensor is not wilfully moved, instead they rely on inevitable random movement between shots to sample at different offsets, then use super-resolution (relying on aliasing) to generate a higher-resolution output image.
-h
-
I was commenting the Hasselblad method, which I thought only did the first (moving the sensor at integer sensel distances so as to cover a full color image using a Bayer color filter)?
According to the Hasselblad H4D-200MS brochure (http://www.hasselbladusa.com/media/2658914/uk_h4d_200ms_v3.pdf):
High precision piezo-electrical actuators control movements of the
sensor in ½ and in one pixel increments. By combining six shots,
offset by a combination of both ½ pixel increments and one pixel
increments, the colors, Red, green and Blue of each point are
obtained with a double resolution in both the X and Y directions.
Therefore, the MS200 uses both of the methods I mentioned, (presumably) two overlapping half pixel increments for denser sampling, and four whole pixel increments to improve color accuracy and avoid demosaicing issues (false color artifacts and loss of resolution).
Cheers,
Bart
-
According to the Hasselblad H4D-200MS brochure (http://www.hasselbladusa.com/media/2658914/uk_h4d_200ms_v3.pdf):
I stand corrected.
-h
-
Here is an article I just finished discussing shooting color with the Achromatic back.
http://brianhirschfeldphotography.com/2013/05/17/shooting-color-with-the-phaseone-iq260-achromatic-prototype/
-
Megavision has a repo system that uses this technique... They also have patents on some of their ideas.
In addition to the filter wheel they also have systems that use colored light sources such as, but not limited to, RGB Leds. A different exposure is done with each light source. Sinar used the color filter technique a long time ago with their tri color shutter.
Many years ago I thought about using the technique with 8x10 BW film but decided it wasn't practical.