Newsgroups: sci.image.processing
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!udel!gatech!howland.reston.ans.net!news.moneng.mei.com!uwm.edu!msunews!harbinger.cc.monash.edu.au!bruce.cs.monash.edu.au!alanf
From: alanf@cs.monash.edu.au (Alan Finlay)
Subject: Re: The term "pixel" and multispectral data, SUMMARY
Message-ID: <alanf.790223505@bruce.cs.monash.edu.au>
Summary: name for a single band component of a multispectral pixel
Keywords: pixel terminology
Sender: news@bruce.cs.monash.edu.au (USENET News System)
Organization: Computer Science, Monash University, Australia
Date: Mon, 16 Jan 1995 02:31:45 GMT
Lines: 111

My original question:
> It seems most consistent to use the term "pixel" to refer to the aggregate
> of all bands at a particular image location (e.g. (red, green, blue) at
> [x,y]). What then do we call the component of a pixel corresponding to a
> single band (e.g. the red value at [x,y])?  In particular I need to apply
> this terminology to name certain procedure parameters which will indicate
> the number of "bits per pixel" in some cases and "bits per ?" in other
> cases.

A summary of individual responses is at the end of this message.  Basically
there is some disagreement about what the term "pixel" should mean.
Of those who agree with my usage above, the terms recommended are
"component" and "sample".  While these are sensible they are too general
for my purpose.  I am considering which of the following to use as a type
name: "pbtype", "pixbase", "pixelbase", "scalarpix".  Where I mean "pixel
base type" and "scalar pixel type" respectively.  One particular difficulty
I have is that my library will use this term to refer to a single component
of a multispectral pixel in some circumstances and the whole pixel value
when the image has only one component.  I thought about defining a new
word, perhaps something begining with the greek prefix "prot" meaning
the basic or fundamental form (e.g. as used in "proton").  However my
knowledge of word formation is not up to it.

If anyone has any preference for one of the four terms:
"pbtype", "pixbase", "pixelbase" and "scalarpix" or anything similar,
please let me know; by EMAIL only, as it is unlikely to interest anyone else.

Here are extracts of some of the responses I received (my comments in <>):

=============================
From jodybooth@aol.com  Sun Jan 15 08:50:08 1995

>In the display engineering area, we have taken to using pixel to refer to
>a 'dot', which can be red, green, or blue.  We use the terms 'color group'
>or 'triad' to refer to the aggregate of three pixels.  I realize that this
>is backwards from most computer terminology.  It may be characteristic of
>color flat panel displays, where the individual red, green, and blue
>pixels are, in fact, addressable.  Pixel is shortened from picture
>element, and whether your 'atomic' unit is a dot or color group depends on
>your perspective on the elements addressability.  So, this suited us quite
>well, but may be confusing in a more general context.

=============================
From park@CS.WM.EDU  Wed Jan 11 16:37:37 1995
Subject: what is a pixel ...

<that wasn't really my question, so I have omitted most of this reply> 
 
>It is important to distinguish between "pixel" and "pixel value".  A pixel is
>a spatial location in an image/scene; a pixel value is the data associated 
>with that spatial location.  For a conventional "mono" image, the pixel value
>...

<Actually "pixel" stands for "picture element" and is a display oriented term>

=============================
From olea@netcom.com  Wed Jan 11 13:29:25 1995

>"Sample" is the term used in TIFF documentation (e.g. bits per sample
>and samples per pixel).

=============================
From: jbm@eos.arc.nasa.gov (Jeffrey Mulligan)
>I don't know about "conventional," but in my own work I use exactly
>the terms you did above:  "component" to refer to a single element,
>and "pixel" to refer to the aggregate.

=============================
From lhszeto@ucdavis.edu  Fri Jan 13 05:27:29 1995

>We work with multispectral images and we usually describe them using the 
>original data terms as not to confuse the reader.  We keep with words 
>like "band" and "pixel"; i.e. "the red band in pixel (x,y)..." or "at 850nm, 
>pixel (x,y) shows..."

>I work in an environmental remote sensing lab.  I am not sure if the 
>above described is appropriate for your field but I think to stay with 
>terms people use in the original data lets reader backtrack to understand 
>derivations in the latter parts of the discussion.

=============================
From: mac@si.sintef.no (Mats Carlin)

>One suggestion is to define an image to be a matrix of pixels,
>where each pixel consists of one or more samples. All pixels in an image
>have the same number of samples.

<mentions TIFF usage of "sample">

=============================
From: perry@netcom.com (Perry West)

>I always believed pixels could be multi-valued, and that single valued 
>pixels (e.g., gray-scale) was just a limited case of multi-valued.
>...
>I have always viewed the pixel as the *atom* of an image in the sense you 
>cannot divide it (spatially).  However, with processing, its value or 
>contents can be modified.

=============================
Reply-To: greg@gkrcc.com

>I've always seen and used the word you did, "component". It's quite
>unspecific, so Hue, IR, or UV components also seem approriate.

=================<end>=======
--
,-------Alan Finlay, email: alanf@bruce.cs.monash.edu.au, phone: (03)565-5225.
|	World Wide Web URL: 	http://www.cs.monash.edu.au:80/~alanf/
`-------Computer Science Dept., Monash University, CLAYTON 3168, Australia.
	========== FRAUDULANT THIRD WORLD DEBT SHOULD BE WRITTEN OFF ========
