[Ilugc] [OT] Images only in Rectangle ?
- From: hari@xxxxxxxxxxx (Srihari Rasmi)
- Date: Tue, 10 Dec 2002 08:43:12 +0530
yes, there is something to add here. the problem is clearly not the
question of rectangular objects or UI-lameness though it surely
is a factor.
Please...the rectangular object _is_ the issue. And its _not_
UI-lameness...its the UI's simplicity. Keeping things simple has always
been the key. Contrary to intuition...a rectangle needs only four
variables i.e., (x_1,y_1),(x_1,y_2),(x_2,y_1),(x_2,y_2). Even a
quadrilateral would need eight variables i.e (x_i,y_i); 0<i<5; i belongs
'Fractally', the pixel is a rectangle. Its easy to build an exact huge
rectangle out of small rectangles. No matter how small the pixels, you
cant build an exact triangle.
[OT] When things come to windows, mouse sensitivity within a window
needs only one parameter specifying the boundary (the horizontal bound
or the vertical) when the mouse is moving in a straight line along _any_
direction...this is very simple. But think about a quadrilateral...you
would need to specify the slope _and_ an initial starting point for the
boundary line...which means...we need two parameters. More
complex...lets stay simple.
the real issue is you are always interpolating a
2d real coordinate system [your image] to a 2d integer coordinate
system [your screen]. the loss of precision here can be offset
cleverly in many ways namely anti-aliasing, shading etc.
I dont see how this is relevant to the topic of rectangular images.
[OT] _Every_ image is sampled (digitised). Even our eye does this.
There are two issues involved here 1) the bit rate..or depth resolution,
and 2) the screen resolution. The more of both...the better.
There is _no_ interpolation anywhere! There is only truncation involved.
For eg., draw a one-pixel thick straight line, with a slope, in an
losslessly compressed or uncompressed format...say .png...and see the
effect of resizing this image in a browser window. This should explain
And Anti-aliasing was brought in because our eyes do a lousy job at
sampling...and because the computer monitor has a fixed screen resolution.
increasing the granularity of what we call a `pixel` is really the
best way to get around such interpolative errors and thats why
as resolution increases your images appear sharper, better and more
"dense". taking this argument , at a more fundamental level its a
problem between analog [nature] and discrete [computer] domains. its
really the reason why we have to "sample" signals to feed into
"digital signal processors" [computers].
One can never know the difference between the analog and discrete. We
always 'see' digital. The eye does a spatial sampling (a frame) and a
time sampling (movie). What we see is a movie played at approx. 30
frames per sec.
So...the real problem is how to make things discrete so that its more
acceptable and pleasing for us (for eg. anti-aliasing).
from your reply i am able to get the message " Till date ,computers can
create images only in rectangular base" is that the meaning ?.
Close. Necessity is the mother of all inventions.
UI gurus have only been thinking about rectangular UI objects. There
hasn't been a need for irregular objects.
Tell me why you would want an irregular image.
is it not possible to make a new format of image where the base is not a
Sure, it is possible. But think of the amount of information you've got
to store about each pixel. The best (and only) approach is to stick with
transparency. Transparency with layering (flat) can be neat.
That is precisely why transparency was brought in.
Do you want an (Textual) Image Viewer ? Please
Is there anything like a textual image viewer??? Its difficult for me to
comprehend how this would work. Any references/links?
Now...coming to images...as to why things are easy when they are
rectangular. They key is this: the encoding of an image is very orderly
when a rectagle is maintained. _Any_ image encoder exploits this.
As mentioned before, if an 'irregular' image is needed, specify the
necessary pixels to have 100% transparency (called alpha in most cases).
I dont think JPEG can do this...any decent lossless compression scheme
like PNG has transparency.
Other related posts: