In article <slr### [at] linuxparkerrfwicom>, 
ron### [at] povrayorg wrote:

> On Wed, 03 May 2000 06:56:52 -0400, Glen Berry wrote:
> > Modern films don't exhibit this effect.
> 
> But some digital cameras seem to, oddly enough.

Ours doesn't...and I have taken several sun-backlit photos with it.

-- 
Christopher James Huff - Personal e-mail: chr### [at] yahoocom
TAG(Technical Assistance Group) e-mail: chr### [at] tagpovrayorg
Personal Web page: http://chrishuff.dhs.org/
TAG Web page: http://tag.povray.org/

Post a reply to this message

In article <slr### [at] linuxparkerrfwicom>, 
ron### [at] povrayorg wrote:

> >> -1 0 0
> >>  0 1 0   * .5 + .5
> >>  0 0 0 
> >
> >The numbers in the matrix above are correct. I'm not sure about the
> >multiplier and offset modifiers. For me, a vaue of one for the
> >multiplier works, and the offset value should be set to one-half of
> >the maximum possible value. I just tested these, and they work.
> 
> A multiplier of 1 can cause clipping at the ends of the range.  The
> two worst-case values of the matrix are -1 (when the upper-left pixel
> is white and the center is black) and 1 (vice-versa.)  To get this range
> into the 0..1 range requires multiplying by .5 and adding .5, as I did.

This sounds like it would best be done as a simple macro using 
blur_matrix(which should probably be renamed to "convolution_matrix", 
since it can do more than blurring and it seems to be the common name).


> A better way to increase contrast would be to apply the
> filter as I've specified it, then apply an S-curve-shaped correction
> to stretch the midrange.

I have been thinking about this...is there a "standard" curve used for 
this, or should I just come up with something?

-- 
Christopher James Huff - Personal e-mail: chr### [at] yahoocom
TAG(Technical Assistance Group) e-mail: chr### [at] tagpovrayorg
Personal Web page: http://chrishuff.dhs.org/
TAG Web page: http://tag.povray.org/

Post a reply to this message

In article <ie71hscr585hiohhurfjkp8bq9pe1ppnt5@4ax.com>, Peter Popov 
<pet### [at] usanet> wrote:

> www.gimp.org

Uh, right. Thanks. :-)

-- 
Christopher James Huff - Personal e-mail: chr### [at] yahoocom
TAG(Technical Assistance Group) e-mail: chr### [at] tagpovrayorg
Personal Web page: http://chrishuff.dhs.org/
TAG Web page: http://tag.povray.org/

Post a reply to this message

On Wed, 03 May 2000 18:10:46 -0500, Chris Huff
<chr### [at] yahoocom> wrote:

>This sounds like it would best be done as a simple macro using 
>blur_matrix(which should probably be renamed to "convolution_matrix", 
>since it can do more than blurring and it seems to be the common name).

If you want, you could also shorten that to simply "convolve" and save
some typing. Since we are referring to a process and not an object,
the verb might also be more appropriate than the noun as well. 

Just an idea...
Glen

( Remove the "7" from 7no### [at] ezwvcom to email me. )

Post a reply to this message

"Chris Huff" <chr### [at] yahoocom> wrote in message
news:chrishuff_99-80DFAD.19253627042000@news.povray.org...
> Does anyone have any suggestions for additional filters?

A GTran filter - makes your images look like Giles did them.

Post a reply to this message

On Thu, 4 May 2000 17:12:33 +0100, "Tom Melly" <tom### [at] tomandluf9couk>
wrote:

>"Chris Huff" <chr### [at] yahoocom> wrote in message
>news:chrishuff_99-80DFAD.19253627042000@news.povray.org...
>> Does anyone have any suggestions for additional filters?
>
>A GTran filter - makes your images look like Giles did them.

While we are wishing for the impossible, I want the GTran+10 filter.
It makes your images look like a Giles rendering from 10 years into
the future. (He might be excellent now, but just imagine how great he
will be 10 years from now.) As a small side effect, it also speeds
your rendering time an amount equivalent to the speed increase of
computers in the next 10 years.

Later,
Glen Berry

( Remove the "7" from 7no### [at] ezwvcom to email me. )

Post a reply to this message

On Wed, 03 May 2000 18:10:46 -0500, Chris Huff
<chr### [at] yahoocom> wrote:

>> A better way to increase contrast would be to apply the
>> filter as I've specified it, then apply an S-curve-shaped correction
>> to stretch the midrange.
>
>I have been thinking about this...is there a "standard" curve used for 
>this, or should I just come up with something?

For what it's worth, here is one example of an S-curve:

    Output = 0.5-(cos(Input*pi)/2)

For this to work, both input and output need to be constrained between
0 and 1. It should work with other ranges of values with appropriate
scaling. The angular unit is expressed in radians, in this example.

This generates an S-Curve, but I'm not sure of an easy way to add
additional parameters and make it more flexible, without losing the
"S" shape.

This curve is also very similar to a generic representation of the
response curve of photgraphic film. It would need some modification to
work as a film response emulator, a subject I raised earlier. The
input range would need to accept those unclipped float values, and the
ouptput would need to be constrained to a range suitable for the
output file format. I'm not yet sure how to add the extra parameters
to do this properly.

Later,
Glen Berry

Later,
Glen Berry

( Remove the "7" from 7no### [at] ezwvcom to email me. )

Post a reply to this message

In article <UcQROZw2W1joV6Ajm0U7uS80aIp5@4ax.com>, Glen Berry 
<7no### [at] ezwvcom> wrote:

> For what it's worth, here is one example of an S-curve:
> 
>     Output = 0.5-(cos(Input*pi)/2)
> 
> For this to work, both input and output need to be constrained between
> 0 and 1. It should work with other ranges of values with appropriate
> scaling. The angular unit is expressed in radians, in this example.
> 
> This generates an S-Curve, but I'm not sure of an easy way to add
> additional parameters and make it more flexible, without losing the
> "S" shape.

This is certainly "a" S-curve, but I wanted to know if there is a 
formula for one which is considered the "standard", official or not...a 
possibly better solution would be a polynomial curve, or something 
similar to what I posted in .binaries.images a while back, where you 
could change it from an S-curve to a straight line to a 
backwards-S-curve, by modifying the exponent. This also makes a S shape, 
and the version I posted works with "out of range" values.


> This curve is also very similar to a generic representation of the
> response curve of photgraphic film. It would need some modification to
> work as a film response emulator, a subject I raised earlier.

This sounds like it would best be done with the "curves" filter I am 
planning, which takes a 3D spline or a color_map for the curve...

-- 
Christopher James Huff - Personal e-mail: chr### [at] yahoocom
TAG(Technical Assistance Group) e-mail: chr### [at] tagpovrayorg
Personal Web page: http://chrishuff.dhs.org/
TAG Web page: http://tag.povray.org/

Post a reply to this message

On Thu, 04 May 2000 16:48:44 -0500, Chris Huff
<chr### [at] yahoocom> wrote:

>> This curve is also very similar to a generic representation of the
>> response curve of photgraphic film. It would need some modification to
>> work as a film response emulator, a subject I raised earlier.
>
>This sounds like it would best be done with the "curves" filter I am 
>planning, which takes a 3D spline or a color_map for the curve...

Splines sound like a good idea. I wasn't sure how elaborate you were
wanting to get. It would be really nice to be able to enter a series
of values (input vs output levels), and have the software generate a
smooth response curve from them.

Later,
Glen Berry

( Remove the "7" from 7no### [at] ezwvcom to email me. )

Post a reply to this message

In article <FUMSOZRek52TE5mBLKHN7VYeCs19@4ax.com>, Glen Berry 
<7no### [at] ezwvcom> wrote:

> Splines sound like a good idea. I wasn't sure how elaborate you were
> wanting to get. It would be really nice to be able to enter a series
> of values (input vs output levels), and have the software generate a
> smooth response curve from them.

That is what I am planning...but I am debating whether to use an 
internal spline(probably a cosine spline, maybe an optional cubic 
spline), or to use a color_map(easier, multiple "film types" can be 
declared in an include file and used in the "curves" filter, but I 
suspect only uses linear interpolation.).
Maybe a "blend_type" option should be added to _maps to switch between 
cosine, linear, cubic, etc. This would solve the problem with using 
color_maps, and make anything which uses color_maps, texture_maps, 
normal_maps, etc more flexible as well.

-- 
Christopher James Huff - Personal e-mail: chr### [at] yahoocom
TAG(Technical Assistance Group) e-mail: chr### [at] tagpovrayorg
Personal Web page: http://chrishuff.dhs.org/
TAG Web page: http://tag.povray.org/

Post a reply to this message