shed skin 0.9.1

[srepmub]

hi all,

I'm planning on doing a shed skin 0.9.1 "maintenance" release during
the weekend. please let me know about any problems you haven't
reported yet, or if you have/know of interesting programs that I might
play with, so I can try and pack a few more things into the release.

the release notes are already up:

http://code.google.com/p/shedskin/wiki/releasenotes

thanks to all those that contributed to this release!

in the meantime, it looks like shed skin 0.9 has also made it into the
debian package repository, thanks to the efforts and persistence of
paul boddie..! :-)

my 3d-printer project is almost in a state where I can start
allocating more time for shed skin and other hobbies again. hopefully
also during the weekend (while running shed skin tests) I will be able
to print my first object.. :P

thanks,
mark.

[srepmub]

while looking for a new example to add for 0.9.1, I came across this
nice image scaling algorithm:

http://www.hiend3d.com/hq2x.html

after a few hours of shooting myself in the foot, I finally converted
this algorithm (well, not really an algorithm but 256 expert
decisions.. ;-)) to python, and added it to shedskin/examples.

mark.

[srepmub]

I just released 0.9.1. please let me know about any problems, before I
announce it more widely..

[Enzo Erbano]

Hi, 

Where you found the source code?

2012/1/14 srepmub <mark....@gmail.com>

--
You received this message because you are subscribed to the Google Groups "shedskin-discuss" group.
To post to this group, send email to shedskin...@googlegroups.com.
To unsubscribe from this group, send email to shedskin-discu...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/shedskin-discuss?hl=en.

[Mark Dufour]

>
> Where you found the source code?

hmm, there seem to be two versions of the website. this one has the source code:

http://spacy51.sp.funpic.de/hq_filters/hq2x.html

mark.
--
http://www.youtube.com/watch?v=E6LsfnBmdnk

[Enzo Erbano]

And the 16xFSAA ?

There is a reference of a demo of it, I cant find anything.... this site is too messy....

Is the 16xFSAA used in hq2x itself?

2012/1/14 Mark Dufour <mark....@gmail.com>

[Mark Dufour]

hi enzo,

> And the 16xFSAA ?

I think this one is relatively simple. just take 4x4 monochrome
pixels, count how many are 'on' and that's the value for the resulting
pixel.

> Is the 16xFSAA used in hq2x itself?

no, the two work in opposite directions. hq2x makes an image larger,
and fsaa makes it smaller.

thanks,
mark.
--
http://www.youtube.com/watch?v=E6LsfnBmdnk

[Enzo Erbano]

Thanks for the explanation,

Maybe I got confused, I thought it supersample a image 16x [16 times the actual resolution],

it means increase in the resolution, then, collect data points at a greater resolution, then 

combine pixels and downsample in a way to solve artifacts, jaggies... etc.... 

I know this technique is very CPU expensive, but I will be going to use it for non-realtime 

photo processing....

Also there are dozens of antialiasing techniques.... but, FSAA, to be more exact, the 

SSAA [super sampling antialiasing] variant, is considered the one with highest quality....

but, I cant find any demonstration or any interesting document about the subject....

I also know that there is an enhanced version of SSAA, from Nvidia, called Sparse Grid 

Supersampling [SGSSAA], that is even better, but, since it is directly implemented 

in hardware, and it depends on the hardware, it is impossible to deliver the very 

same results of AA to all people that would use the software.... it is the reason that

I am very interested in implement it on software....

Anyway, maybe I am not in the right place to talk about this... hehehehehehe

Best regards.

2012/1/16 Mark Dufour <mark....@gmail.com>

mark.
--
http://www.youtube.com/watch?v=E6LsfnBmdnk

[Mark Dufour]

> Anyway, maybe I am not in the right place to talk about this... hehehehehehe

well, it might result in a nice new example program.. ;-)

mark.
--
http://www.youtube.com/watch?v=E6LsfnBmdnk

[Enzo Erbano]

Indeed!

This subject is quite interesting, and would be a nice one to get in shedskin....

If you get interested in the subject, there is one other very cheap AA known as FXAA.... 

that works as post-processing, have a nice quality, is independent than hardware

 and software.... the processing is only based on color....

Just take a look on this page  http://timothylottes.blogspot.com/2011/12/fxaa-40-development-update-stills.html 

Notice in the examples that the edges are very smooth, you cant notice jaggies and artifacts.... 

but it does not make the image blurry like others AA.... and the images are have a very nice 

quality, but, I dont think it is better than the other I mentioned.... but it is very very easy to 

integrate in any engine... and works in realtime...

The sources are here...

http://timothylottes.blogspot.com/2011/07/fixed-few-typos-in-fxaa-311.html

2012/1/16 Mark Dufour <mark....@gmail.com>

mark.
--
http://www.youtube.com/watch?v=E6LsfnBmdnk

[Mark Dufour]

hi enzo,

sure this is fascinating stuff.. unfortunately I can't get too deep
into other projects at the moment though. let me know if you have/find
some working python code that we could (almost) run through shedskin.

I implemented a simple "snappy" decompressor last week, but it was too
small to add to examples/. perhaps a compressor would be more
interesting, but I haven't looked into this yet.

thanks,
mark.

--
http://www.youtube.com/watch?v=E6LsfnBmdnk

[Enzo Erbano]

Hi Mark,

Unfortunately I not found any code in python, only in Java (2 versions
of supersampling, but it would be necessary to isolate only the code
for supersampling because it is messed with other codes, and then
translate to python) or C/C++, or even in shading language (it is the
case of FXAA)...

I know that you are busy to other projects... and cant get more things
to develop.... I understand, but.... the FXAA is not a big code.... most
of the file is purely comments explaining how it works! Near, 6 to 10 lines
of comments for each line of code, and also, there are many variations
of the same code for many targets, so, isolating only the interesting
version, the one with highest quality.... there is only 500 lines of code...

By the way, I think worth porting FXAA, as it is from Nvidia, and is
becoming very popular... and is fast with a good quality....

The case of supersampling, using poisson disc for distribution, well, it
is the most high quality AA I found of dozens techniques, but the
processing it takes is really overkill.... and I bet no one will use this...

It is the direct link to the FXAA code...
https://docs.google.com/leaf?id=0B2manFVVrzQAOWUwODM1YmQtNmEwOC00ZDBlLThlYzMtYzE4ZDM1ZjRlMzMx&sort=name&layout=list&num=50

2012/1/20 Mark Dufour <mark....@gmail.com>: