New wave colour processing

[papa v2.0]

Hi guys

Heres some stuff that ive been working on.

hiweb.co.uk
I now have my paper published at ICIS Beijing 2010 (International Conference  on Imaging Science).

So I can let you have look at things to come

I have been working on introducing colour appearance into a RAW imaging work flow using CIECAM02.
I have posted some of the first  images produced by the project if you would like have a deek at them.

Basically I have moved away from the current tends in imaging  and instead of rendering from Colormetirc scene estimates I have rendered from Appearance estimates.

This is a software implementation before the user get their grubby paws on the image. At the mo its a sRGB workflow but can be adapted.
I have bypassed Prophoto completely. Their are certain advantages of going to prophoto then to sRGB but ill discuss later.

I have a series of images  the first is from appearance  and the second is from the nikon jpeg.
The RAW file i used is the same used by the camera to generate the jpeg. i used all the same settings as the camera as per white point exposure etc. there was no colour processing in in photoshop so what you see is  the image straight afer a basic clipping to the sRGB gamut.

That said I think with a bit of preferential rendering i can vastly improve these initial images.

Any feed back should be interesting!

ps these image have been take with a colour science point of view as opposed to a potographic award point of view.

many thanks Iain

PPS

Ross County for the CUP!!!!

[Guillermo Luijk]

I find your images quite warmer, not sure if just a difference in white balance (although you said you used the same as camera, so the difference should come from your approach). What is exactly the basis of your method with respect to the common way?

Also deep shadows are a bit lifted in your images with respect to the original JPEG files, but I guess this could be explained because of a different contrast curve applied when creating camera's JPEG. I find some partial saturation in the highlights in your images, could have the same explanation as the shadows.

[papa v2.0]

I find your images quite warmer, not sure if just a difference in white balance (although you said you used the same as camera, so the difference should come from your approach). What is exactly the basis of your method with respect to the common way?

Also deep shadows are a bit lifted in your images with respect to the original JPEG files, but I guess this could be explained because of a different contrast curve applied when creating camera's JPEG. I find some partial saturation in the highlights in your images, could have the same explanation as the shadows.

Hi

what i am doing is to develop a situation were the image is rendered from appearance coordinates instead of colorimetric coordinates.
The matrix i have used is developed in an environment where the errors are minimised in JMh space as opposed to a calorimetric space
ie conventional L*a*b* (yes its a perceptual space!, but not a good one) or XYZ  or RGB non linear .

These images are just the first attempts. I did not want to introduce and individual rendering, but to look at what CIECAM02  can achieve BEFORE  preferential rendering, if you can see where Im coming from.
I am interested in your blending of RAW files and thats my next experiment!! ZERO NOISE etc

They are not perfect by all means, but i have not really done any post rendering on the files yet.

Iain

[George Machen]

How about a link to your paper, so we can read it and better understand your theory?

[Tim Lookingbill]

Checked out your Appearance Processed rendered version on that site and can't really see how this is an improvement over just beefing up saturation on the original Nikon rendered jpegs which with some look more like the way reality APPEARS to my eyes.

From what visual reference are you basing this Appearance Process version on? Who sees images that way? What is the purpose of making them appear that way?

I find some of them to be an improvement and others appearing as if too much of a yellow to reddish orange filter was placed over the image. But then the Nikon renderings have a color filter appearance as well, just more neutral-ish blue and desaturated. Beefing up saturation somewhat separates the spectral reflectance properties inherent in each object not seen prior to increasing saturation.

This is what I'ld hoped you'ld address in your Appearance Process approach; an ability to separate and amplify different hue variants between cool and warm spectral reflectance appearances according to surfaces brightly lit against those in shadow in an attempt to eliminate this filter appearance. It's the same effect similar to what can be done using Adobe Camera Raw's Split Tone panel which doesn't work all the time with some images. Some of your images almost accomplishes this but there's still an overall filtered feel to them. Is this intentional?

Master landscape and portrait painters 100's of years ago came up with this approach to their renderings because they tackled image making by seeing and reconstructing the scene on canvas object by object similar to constructing a jigsaw puzzle or color mosaic. They disregarded the actual filtered appearance of color temperature's affect on the overall appearance of objects lit in the scene and only concentrated on rendering hues that looked homogenous next to each other but still made the image look real.

I'm glad someone's trying to come up with a new color appearance model. Keep up the good work.