Re: pictogram's Incamera
Re: pictogram's Incamera
- Subject: Re: pictogram's Incamera
- From: Marco Ugolini <email@hidden>
- Date: Mon, 29 Aug 2005 10:33:59 -0700
Title: Re: pictogram's Incamera
In a message dated Mon, 29 Aug 2005 08:30:04, Eugene Appert wrote:
Because of my experience with these other profiles coupled with my understanding (crude at best) of Fraser, Murphy, Bunting I assumed that the sample points of a target were used to deduce primaries from which a virtual model was generated. Therefore the test of a profile would be to turn it back on the target that generated it and measure the variances.
The way I understand it (and others should correct me here if I am mistaken), the profile is not built around the primaries, although it measures those too. Instead, it measures ALL colors in the test chart and builds a lookup table in which those colors are the principal nodes, and where intermediate colors (the ones in between, those not directly measured) are interpolated according to "secret sauce" methods adopted by each profile software manufacturer. Besides interpreting the measured color values correctly, this issue of interpolation is a crucial one for a profile's accuracy and overall validity.
Testing the validity of a profile by reading the colorimetric values of colors after the profile is applied and used, say, for printing to a linearized and profiled inkjet, is a variant of what is known as "round-trip test": ideally, if the profile is top-notch and the output device can reproduce all colors in the original, the values at the end of the process closely match those at the beginning, though even in the best cases there is always a degree of difference. The difference in values is calculated using the Delta-E method (which comes in several versions), either as a statistical mean of all colors, or groups of them, or individual ones.
The ease at which Incamera produced perfect results before my eyes forced me to realise that I had overly complicated the procedure. It was simply writing the known lab values beside the control signals for the 24 sample points. Now that I know that this is all it does I am surprised input-profiling software isn’t free at PC world.
Others may intervene and straighten me out, but I don't think that this is what the software is doing here in Incamera (i.e., recognizing the separate 24 sample colors, then substituting for each its own correct Lab values: that would be an artificial intelligence feature that I doubt is present in that software). As I said before, it seems more likely that some happy coincidence is taking place, whereby the standard profile (applied to the image by default by the image capture software) interprets the chart values correctly because, possibly, it so happens to be a valid profile for the particular camera and situation used in capturing the image.
--------------
Marco Ugolini
Mill Valley, CA
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden