Re: Euclidean distances in L*a*b* - problem!

In article <1120799928.513330.147090@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
kalfholo <mdmcgr@xxxxxxxxx> wrote:
>The relationship is the following. To minimize the data storage
>requirement and maximize the preservation of visual detail, we should
>store images in a perceptually uniform format.

That statement may be true when storage costs are very high. If you
can afford 8-bit/ch tiff (with whatever encoding you use) you can in
many situations just as well afford 16-bit/ch linear storage.

I guess it would be nice to have a jpeg version that can deal with LAB.

>In some cases it's
>convenient to perform image operations directly in that format. For
>example one can do a tone correction by manipulating L* with minimal
>effect on hue and chrominance--within the limitation of gamut of
>course.

Assuming that with 'tone correction' you mean changing the brightness,
you can do that in just about every other commonly used space (both linear
and gamma corrected) as well. I don't see the advantage.

>The crank position is that linear data _is_ perceptually linear
>and that all image processing operations should be performed only in
>linear space.

I think there may be some truth in the second part of the statement.
('should' is not as strong as 'can').

Lots of operations related to the underlying physics and math of light
can only be understood (easily) in linear space. What kind of MTF
do you get when to perform sinc based interpolation in LAB?

>You are quite right that _some_ operations do need to be
>done in linear space--matrix operations, white point correction etc.
>But in many cases but not all, it's not hard to do this with images in
>non-linear space with lookup tables. They are calculated with floating
>point math which effectively passes through linear space but input and
>output are non-linear.

I can't see the advantage of doing complicated operations in LAB using
lookup tables, when you can just as well convert first to linear space,
perform the operations and then convert back.

Assuming that monitors are profiled in addition to being calibrated,
you may want to do a matrix multiplication before displaying data anyhow.


--
That was it. Done. The faulty Monk was turned out into the desert where it
could believe what it liked, including the idea that it had been hard done
by. It was allowed to keep its horse, since horses were so cheap to make.
-- Douglas Adams in Dirk Gently's Holistic Detective Agency
.

Relevant Pages

  • Re: Is camera response needed for HDR based upon multiple exposures ?
    ... In many papers it is suggested to recover first the non linear ... different exposures. ... mixing of the images is carried out in that domain. ... maximizes signal to noise ratio under the assumption of an additive ...
    (sci.image.processing)
  • Re: Scientifically Based Presharpening for Enlargement
    ... Your starting images should be uncompressed and unsharpened. ... interpolation, which is familiar to many people, and doesn't ... You might address why the sharpening should be a preprocessing step. ... the sharpening is a linear filter, ...
    (sci.image.processing)
  • Re: linear regression results approximates to the mean of Y
    ... Linear Regression is used to estimate human age when given a face ... I am not very sure if there's linear realtion between the X and Y. So ... I have 25 images for each age, and the age range is from 20 to 60. ...
    (sci.stat.math)
  • Re: Linear and non-linear brightness scaling
    ... All images are stored by linear data with 16 bits per ... Convert linear 16 bpc data into 16 bpc inverse ... analog video signals to the monitor. ... All images are still stored by 8 bpc. ...
    (sci.engr.color)
  • Re: skew symmetric map/comples geometry
    ... On 21.01.2006 15:28, singau wrote: ... > Let V be a real linear space with inproduct S, ... > and complex structure J can be viewed a a complex linear space with ...
    (sci.math)