Auto-processing to a Preferred color reproduction

[PeterLange]

This early patent by Kodak suggests that there would be higher rules, shifting colors around in 3D, to produce a preferred color reproduction - which even can be automated:

http://www.freepatentsonline.com/6791716.pdf
http://www.freepatentsonline.com/6791716.html
(pdf text starts at p. 20; Hue manipulations p. 25, Chroma enhancements and Lightness scaling p. 28)

What do you think?

Peter

--

[Jonathan Wienke]

You can accomplish the same thing (any color manipulation) with a "tweaked" camera profile. Instead of shooting the color reference chart directly, shoot the reference chart with the film whose "look" you want to emulate, make a print of the film shot of the color chart, shoot the print with your digital camera instead of the original chart and use the film print shot to create the tweaked camera profile. To duplicate the film "look", process the RAW normally, convert to the "tweaked" profile, and then assign your normal editing space.

Suppose you have a digital camera camera that outputs Adobe RGB JPEGs (like the Canon 1D-MkII) and you want to make your JPEGs look like Kodak Portra. Simply follow these steps:

1. Get yourself a camera profiling utility like Eye-One Match and the Digital Color Checker SG color reference chart to go with it.

2. Shoot the Color Checker SG with a film camera loaded with Portra, and have a life-sized print made using your customary developing and printing process.

3. Shoot the Color Checker SG with the 1D-MkII, and feed the resulting JPEG into Eye-One Match. Save the resulting profile as "1D-MkII".

4. Shoot the Portra print of the Color CHecker SG with your 1D-MkII, and feed the resulting JPEG into Eye-One Match. Save the resulting profile as "Portra".

5. Shoot a subject with the 1D-MkII. Open the JPEG in Photoshop, and convert to 16-bit mode.

6. Assign the profile "1D-MkII" to the image. This will cancel out any color deviations from the camera's internal JPEG processing.

7. Convert the image to the "Portra" profile.

8. Assign the Adobe RGB profile to the image. It will now have the same color characteristics as an image shot with Portra film, minus the grain structure.

I'm not sure why Kodak deserves a patent on this, unless they were involved in inventing profiles and the math involved in converting from RGB to CIE LAB and back.

[PeterLange]

Well, it seems that there are already digital targets available for a specific (pleasing) rendition. Here’s an interesting procedure by Eric Chan; how to mimic the look of e.g. Capture One with ACR:
http://people.csail.mit.edu/ericchan/dp/acr-color-match/

Referring to the patent, I’ve tried to work through it step-by-step. My impression is that the use of CIE Lab as working space is quite decisive for the resulting look – which may or may not be seen as pleasing…

That said, my opinion on this subject is not really settled.  For example, just linearize ACR by setting Shadows, Brightness, Contrast and Saturation to zero.  The resulting ‘colorimetric’ rendition will most likely look dark & dull with real-world images for reason of dynamic range compression (scene to monitor).  That’s clear, but which tweaks & transformations would reliably lead to a preferred color rendition in the sense of obtaining ‘just a nice image’ without further effort (?).

Peter

--

[bjanes]

You can accomplish the same thing (any color manipulation) with a "tweaked" camera profile. 

I'm not sure why Kodak deserves a patent on this, unless they were involved in inventing profiles and the math involved in converting from RGB to CIE LAB and back.
[a href=\"index.php?act=findpost&pid=86902\"][{POST_SNAPBACK}][/a]

To patent a process, the process must be novel, non-obvious, and have some utility. In addition, laws of nature are not patentable.

As I understand ICC profiles, they are not adaptive. That is they use look-up tables to map input to output without taking into consideration the content of the image. The matrix conversions are also not adaptive and are purely mathematical.

On the other hand, the Kodak claim is adaptive.  It uses matrix and lookup transforms to shift the hues of memory colors (skin, foliage, blue sky, etc) in a particular fashion. It also maps contrast and luminosity to achieve a more pleasing appearance, rather than representing a literal interpretation of the scene. Saturation is also mapped. I do not think these effects can be done by an ICC lookup table or matrix transform. If these adaptive algorithms meet the above criteria, they could be patentable.

[jani]

To patent a process, the process must be novel, non-obvious, and have some utility. In addition, laws of nature are not patentable.

Yes, that is the theory, yet patents are granted on old, obvious methods quite often.

Don't trust the patent examiners to know one double-linked list from another, or how playful cats are.

[bjanes]

Yes, that is the theory, yet patents are granted on old, obvious methods quite often.

Don't trust the patent examiners to know one double-linked list from another, or how playful cats are.
[{POST_SNAPBACK}][/a]

A good point, but patents that do meet the above criteria can be challenged in court. Of course, the legal costs of a challenge raise the cost of entry and normally are mounted by companies with deep pockets. I agree that some process patents that are granted by the patent office are absurd, such as the one click ordering offered by Amazon.com.

[a href=\"http://www.nytimes.com/library/magazine/home/20000312mag-patents.html]http://www.nytimes.com/library/magazine/ho...ag-patents.html[/url]

Bill

[Jonathan Wienke]

As I understand ICC profiles, they are not adaptive. That is they use look-up tables to map input to output without taking into consideration the content of the image. The matrix conversions are also not adaptive and are purely mathematical.

That's only true of absolute colorimetric conversion. If you use perceptual conversion, the image colors are manipulated to fit into the destination color space, which sometimes causes hue shifts. The adaptation may be based on the relationship between the input color space to the output color space instead of the image data, but it's not an absolute process, and there are variations in how it is done from one implementation to another.

[bjanes]

That's only true of absolute colorimetric conversion. If you use perceptual conversion, the image colors are manipulated to fit into the destination color space, which sometimes causes hue shifts. The adaptation may be based on the relationship between the input color space to the output color space instead of the image data, but it's not an absolute process, and there are variations in how it is done from one implementation to another.
[a href=\"index.php?act=findpost&pid=87135\"][{POST_SNAPBACK}][/a]

Perceptual colorimetric conversion applies a predetermined amount of compression to the image, and does not take the contents of the image into account. Therefore, it is not adaptive. The variations you mention are predetermined in advance according to the preferences of the maker.

A smart perceptual colorimetric conversion process is badly needed and I would think that a novel process could be patented.

Bill

[jani]

A good point, but patents that do meet the above criteria can be challenged in court.

Any patent can be challenged in court.

Whether such a challenge is successful, however, does depend at least partially on those criteria.

Of course, the legal costs of a challenge raise the cost of entry and normally are mounted by companies with deep pockets.

As is the cost of applying for patents. It's not something for Joe Average.

These days, if you have an invention that you think needs patent protection, you must be prepared to spend at least around USD 100,000 to register your claim with the relevant patent and trademark offices around the globe, and probably more if you include legal aid. In addition, you must be prepared to spend effort, time and money in defending your legal rights.

In practice, patents are for the well-off.

Whether that in itself is bad or not is a even more off-topic, so I'll stop bothering people with this topic right now, and I also apologize to those who have found my posts bothersome.