38980

神经网络昆斯林的新闻组分类2006

Xref: cantaloupe.srv.cs.cmu.edu alt.binaries.pictures.supermodels:946 alt.graphics.pixutils:5395 comp.graphics:38980 alt.binaries.pictures.erotica.d:15191 alt.binaries.pictures.d:8917
Path: cantaloupe.srv.cs.cmu.edu!magnesium.club.cc.cmu.edu!news.sei.cmu.edu!cis.ohio-state.edu!zaphod.mps.ohio-state.edu!cs.utexas.edu!uunet!mcsun!news.funet.fi!butler.cc.tut.fi!lehtori.cc.tut.fi!not-for-mail
From: [e-mail] (Kouhia Juhana)
Newsgroups: alt.binaries.pictures.supermodels,alt.graphics.pixutils,comp.graphics,alt.binaries.pictures.erotica.d,alt.binaries.pictures.d
Subject: Re: XV problems
Date: 30 Apr 1993 10:01:38 +0300
Organization: Tampere University of Technology
Lines: 100
Distribution: world
Message-ID: <1rqisi$[e-mail]>
References: <[e-mail]> <1rohjc$[e-mail]> <[e-mail]>
NNTP-Posting-Host: cc.tut.fi

In article <[e-mail]>
[e-mail] writes:
>
>In article <1rohjc$[e-mail]>, [e-mail] (Kouhia
>Juhana) writes:
>
>>I wrote something about making color modifications quickly
>>with 8bit quantized images and only at the saving the image to file
>>process we have to make the modifications to the 24bit image.
>>This makes sense, because the main use of XV is only viewing images.
>>
>>Doing many changes to image, we should keep all modifications
>>in a buffer; and then before making the operations to 24bit image,
>>we should simplify the operation list for unnecessary operations.
>>
>Think about what you are saying here. The 24 bit image is quantised down to 8
>bits so many 'similar' colours are mapped onto a single palette colour. This
>colour gets modified in fairly arbitrary ways. You then want to apply these
>modifications back to the 24 bit file, so you have to find which
>colours mapped to this one palette colour.

I suppose you don't know what about we have discussed.
We discussed about error(s) in XV 2.21 which shows images only as 8bit,
and my suggestion above works perfectly with it.

So far I have seen a colormap editing window in XV -- that is, there
must be a colormap anyway. The problems you present are exist anyway,
and I didn't tried to solve them at all, because I would not make such
problems to my programs in the first place.

Gamma and color corrections are easily done to 24bit image
as I presented. There's no need make tricks from 8bit/quantized image
back to 24 bit image.


>>>How would you suggest doing colour editing on a 24 bit file? How
>>>would you group 'related' colours to edit them together? Only global
>>>changes could be done unless the software were very different and
>>>much more complicated.

Ok, you're writing about situation that user want edit images as 24bit
and user want edit individual colors -- your questions, by the way,
jumps off the discussion a bit.

My solution doesn't work, because there's no colormap withing real 24bit
image -- you see, user see 24bit image; going back to 8bit is silly.


About changing individual colors in 8bit/quantized/rasterized image:
changing individual colors in colormap is useless in most
cases if the image is quantized and rasterized -- small change may
make serious errors to anywhere in the image.
XV allows this feature, but I don't recommend to use it with the
mentioned type images.

Moreover, XV is not a paint program; you can only make those global
changes. In full 24bit XV, changing individual colors sounds like
paint program job.
If person have 8bit screen, there's need for tricks to get the
original 24bit image modified. Because user don't see full 24bit
image, there's need to make approximations and it is not possible to
modify individual colors but individual pixels or pixel groups (if
image is rasterized). To select indiavidual color, there could be 7x7
cursor window which shows true color image in cursor window area --
selecting individual color is possible from that.

Ok, I don't have thought very much 24bit painting programs, never seen
such in good view and are not planned to make such. Not to mention
24bit painting program in 8bit screen...


>Yes again. What *is* (was?) wrong with xv?

It saved 8bit/quantized/rasterized images as 24bit jpegs; jpeg is not
designed for that.
Also, human expect that 24bit will be saved as 24bit image; say,
person would like to crop part of the image and save it, then it is
expected that the image still is the same. So, XV were designed
without thinking about human interface and how human expect the
program work -- design error.

I have heard XV were designed first for 8bit images/files, but
it were not good idea to take full 24bit images without making
major change to the original design.

So, even all screen images are 8bit, the processed images and saved
images could have been 24bit very easily, instead of 8bit.

Before anybody will make a note: yes, I may as well make a lift where
'up' means that the lift goes down and 'down' means that the lift goes
up, and put a note on this design solution to the manuals -- however,
even the manuals tells the correct situation, it doesn't solve the problem.
(Americans: the lift is just an example :)


Well, my text may be a bit hard reading, hopefully you suggeeded to
read it.


Juhana Kouhia