Open Photography Forums  
HOME FORUMS NEWS FAQ SEARCH

Go Back   Open Photography Forums > Color Management Discussion > CM Theory and Practice

CM Theory and Practice Profiles, color spaces, perception, science.

Reply
 
Thread Tools Display Modes
  #1  
Old July 14th, 2011, 09:51 AM
Doug Kerr Doug Kerr is offline
Senior Member
 
Join Date: May 2006
Location: Alamogordo, New Mexico, USA
Posts: 8,608
Default Orthonormality in coordinate systems

As I reviewed the work of Sergey Bezryadin with respect to the color space used in the Unified Color, I ran repeatedly into the matter of the orthonormality of a coordinate system.

I did a little research into the formalities of this, and then of course had to dispose of some paradoxes that brought to mind.

Here is how it looks to me now.

************
I will of course paraphrase here to keep the matter of understanding moving forward.

A coordinate system may be considered orthonormal if:

Its axes are all orthogonal (for a two- or three- dimensional system, visualized as physical, that means that they are mutually at right angles)

Its axes all have the same "scale"

We can perhaps best understand this if we consider an actual physical coordinate system of three dimensions - for example, as we might use to describe a point in space with respect to some reference point on my optical laboratory bench.

Suppose we define the axes thus (the intent being that they in fact all be at right angles to one another in the physical, geometric source. Perhaps X is to the East, Y to the North, and Z up.

Then we say that, for all of them, one unit of the numerical coordinates equals one meter of physical space.

This is an orthonormal coordinate system. (Coordinate system A).

Now suppose we create another coordinate system where X and Y are in fact in a truly horizontal plane (in the directions previously mentioned) by the X axis is perpendicular to a certain tilted surface in my apparatus.

This is no longer an orthonormal coordinate system. (Coordinate system B)

Or suppose that we keep the original mutually-orthogonal axes but for X and Y consider one unit to correspond to one foot, but for Z consider one unit to correspond to 2 feet.

This is no longer an orthonormal coordinate system. (Coordinate system C)

An impractical distinction between the first coordinate system and the other two is this (a matter than is of great importance in the field in which I recently encountered the concept).


Euclidean distance

The Euclidean distance between two set of values in a coordinate system is an "abstract distance" between the points representing them. If one point is our origin (X,Y,Z=0,0,0), and the other is at A, B, C (this is in abstract units of the axes, not physical feet), then the Euclidean distance between those sets of values is given by:

S=sqrt(A^2+B^2+C^2)

Now, in an orthonormal coordinate system (in physical space), the Euclidean distance between two sets of numbers is the same as the actual physical, spatial distance between the points represented by those sets of numbers (or is related to it by a constant factor, depending on the actual "scale" we use for all three axes).

If the coordinate system is not orthonormal, that is not true.

Related coordinate systems

Suppose we take our first coordinate system (which we have declared to be orthonormal) and just twist it, intact, to another orientation. This means that two or perhaps all three axes now point in different directions that before (perhaps we have aligned it with some frame in our optical apparatus).

This new coordinate system will be orthonormal. (Coordinate system D)

As an intuitive test of that, consider two points in our physical apparatus that are exactly 1.000 inch apart (and let's assume that the scale of all our axes is 1.000 unit equals 1.000 inch.

Now in our first coordinate system, system A (orthonormal), if we take the ordinate value sets that describe these two points and determine the Euclidean distance between them, we will get 1.000 unit.

If, in coordinate system D (also orthonormal), we take the ordinate value sets that describe these same two points and determine the Euclidean distance between them, we will get 1.000 unit (because the actual distance is unchanged, and the "measuring frame" is unchanged, albeit shifted in orientation).

A non-geometric coordinate system

Of course, we often have coordinate systems that do not relate to the actual physical position of a point, but merely are a way to organize the concept of several attributes of an item. Suppose that, for a population of audio amplifiers, we are concerned with their maximum power output (P), their cost (C), and their mass (M).

We can now consider a "three-dimensional" coordinate system in which the axes are P, C, and M. We will consider one unit along the P axis to be one watt, one unit along the C axis to be one USD, and one unit along the M axis to be one kilogram.

Now, is this an orthonormal coordinate system?

The concept is not really defined (it took a couple of minutes for me to get that). For example, what about the notion that the three axes have to be at right angles to one another. Is the axis of Power at right angles to the axis of Cost? That has no meaning.

Certainly, if we draw (for use in a book) an oblique representation of this three-dimensional coordinate system, we will proceed as if all three axes are mutually at right angles. But that is just a graphic custom.

Is the Euclidean distance between the sets of P,C,M values (in abstract form, shorn of their units of watts, USD, and kg) for two amplifiers consistent with the "actual physical distance between their points"? Well, there is no meaning to "the actual distance between their points".

So we must recognize that the classification of a coordinate system as orthonormal or not is predicated on that coordinate system existing in a coordinate system in which "actual distance" has meaning.

Thus, for coordinate systems that describe the location of a point in physical space (where there is a clear meaning of "actual distance" - we could more it with a tape measure, if it's not too big) we can properly describe possible coordinate systems as orthonormal or non-orthonormal.

But when we have a coordinate system that just "comes out of left field" (like our P, C, M system for describing three properties of audio amplifiers) we cannot properly say whether it is orthonormal or not.

Another outlook is that "Of course it is orthonormal. We think of the three axes as being mutually at right angles, and who is to say that they are not".

I will relate the application of this notion to the work of Bezryadin a little later on.

Best regards,

Doug
Reply With Quote
Reply

Bookmarks

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are Off
[IMG] code is On
HTML code is Off

Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
News: Big Brother has big eyes in the sky! Asher Kelman Breaking News 14 February 25th, 2009 03:00 PM
Skin beauty and color coordinate systems Doug Kerr CM Theory and Practice 0 November 15th, 2008 01:49 PM
Color model vs. color space - part 1 Doug Kerr CM Theory and Practice 1 May 9th, 2008 11:18 AM
Mark Hamburg has left Adobe Systems for a new job at Microsoft! Dave McCaughan Breaking News 2 April 29th, 2008 06:02 AM
Michael Reichman's 2006 assessment of MF Digital Systems, where are we now? Asher Kelman Medium Format & Large Format Cameras 10 October 16th, 2007 07:43 PM


All times are GMT -7. The time now is 07:08 PM.


Posting images or text grants license to OPF, yet of such remain with its creator. Still, all assembled discussion 2006-2017 Asher Kelman (all rights reserved) Posts with new theme or unusual image might be moved/copied to a new thread!