What effect does scaling a camera have on focal_blur?

I have two cameras, and they should both produce the same result. But 
they don't.



// focal blur camera
camera {
   location  <0,0,-5>
   look_at   <0,0,0>
   right     x*image_width/image_height
   up y
   aperture 1.0           // [0...N] larger is narrower depth of field 
(blurrier)
   blur_samples 10        // number of rays per pixel for sampling
   focal_point <0,0,0>    // point that is in focus <X,Y,Z>
   confidence 0.95        // [0...<1] when to move on while sampling 
(smaller is less accurate)
   variance 1/200         // [0...1] how precise to calculate (smaller 
is more accurate)
   scale 2
}

/*
// focal blur camera
camera {
   location  <0,0,-10>
   look_at   <0,0,0>
   right     x*image_width/image_height
   up y
   aperture 1.0           // [0...N] larger is narrower depth of field 
(blurrier)
   blur_samples 10        // number of rays per pixel for sampling
   focal_point <0,0,0>    // point that is in focus <X,Y,Z>
   confidence 0.95        // [0...<1] when to move on while sampling 
(smaller is less accurate)
   variance 1/200         // [0...1] how precise to calculate (smaller 
is more accurate)
}
*/

Post a reply to this message

Also, what if I scale the camera before I add the focal_blur? I am 
having trouble with a scene in which the center never appears in focus 
like I want it to.

:(



On 2/10/2018 7:47 PM, Mike Horvath wrote:
> What effect does scaling a camera have on focal_blur?
> 
> I have two cameras, and they should both produce the same result. But 
> they don't.
> 
> 
> 
> // focal blur camera
> camera {
>    location  <0,0,-5>
>    look_at   <0,0,0>
>    right     x*image_width/image_height
>    up y
>    aperture 1.0           // [0...N] larger is narrower depth of field 
> (blurrier)
>    blur_samples 10        // number of rays per pixel for sampling
>    focal_point <0,0,0>    // point that is in focus <X,Y,Z>
>    confidence 0.95        // [0...<1] when to move on while sampling 
> (smaller is less accurate)
>    variance 1/200         // [0...1] how precise to calculate (smaller 
> is more accurate)
>    scale 2
> }
> 
> /*
> // focal blur camera
> camera {
>    location  <0,0,-10>
>    look_at   <0,0,0>
>    right     x*image_width/image_height
>    up y
>    aperture 1.0           // [0...N] larger is narrower depth of field 
> (blurrier)
>    blur_samples 10        // number of rays per pixel for sampling
>    focal_point <0,0,0>    // point that is in focus <X,Y,Z>
>    confidence 0.95        // [0...<1] when to move on while sampling 
> (smaller is less accurate)
>    variance 1/200         // [0...1] how precise to calculate (smaller 
> is more accurate)
> }
> */

Post a reply to this message

I've been reading this

https://en.wikipedia.org/wiki/Miniature_faking

and have been wondering how to achieve the effect in POV-Ray. Obviously, 
there is no true scale in POV-Ray, so there's no need to "fake" the 
effect. But I don't know how to construct the correct camera.

Any ideas?


Mike

Post a reply to this message

I still don't understand exactly what's happening, but I removed the 
scaling and just increased the numbers directly, and things are working 
out much better.


Mike

Post a reply to this message

On 02/10/2018 10:36 PM, Mike Horvath wrote:
> I've been reading this
> 
> https://en.wikipedia.org/wiki/Miniature_faking
> 
> and have been wondering how to achieve the effect in POV-Ray. Obviously, 
> there is no true scale in POV-Ray, so there's no need to "fake" the 
> effect. But I don't know how to construct the correct camera.
> 
> Any ideas?
> 

See:

http://news.povray.org/povray.binaries.images/thread/%3C57e14051%241%40news.povray.org%3E/

for a technique using a not yet adopted parser change which allowed 
image_maps to be used directly in density {} blocks.

Mention it because though our documentation

http://wiki.povray.org/content/Reference:User_Defined_Pattern

doesn't yet disclose it, the new-ish 3.7.1/3.8 pigment { user_defined { 
} } capability was implemented by Christoph in a way so it works for 
density blocks too. Of course only the R,G,B functions do anything where 
used with density {}.

This means it should be possible to implement the tilt shift technique 
in the first reference by building up functions which grab the r,g,b 
channels via #declare Fn... = function { pigment { image_map {...} } } 
and using those functions in two density { user_defined { } } blocks in 
a density_map. It's a more awkward means than using the image_maps 
directly in the density blocks, but it should work.

The technique uses media blending between two densities. One blurred via 
turbulence and one not.

Bill P.

Post a reply to this message

Am 11.02.2018 um 04:36 schrieb Mike Horvath:
> I've been reading this
> 
> https://en.wikipedia.org/wiki/Miniature_faking
> 
> and have been wondering how to achieve the effect in POV-Ray. Obviously,
> there is no true scale in POV-Ray, so there's no need to "fake" the
> effect. But I don't know how to construct the correct camera.
> 
> Any ideas?

Yep: Crank up the focal blur `aperture` setting like mad.

Post a reply to this message

William F Pokorny <ano### [at] anonymousorg> wrote:

>
> ..for a technique using a not yet adopted parser change which allowed
> image_maps to be used directly in density {} blocks.
>
That is *definitely* an interesting and useful idea (as is your current
workaround.) Any idea if this image_map thing is going to be included in a newer
v3.8.x alpha or beta?

Post a reply to this message

On 2/11/2018 10:14 AM, clipka wrote:
> Am 11.02.2018 um 04:36 schrieb Mike Horvath:
>> I've been reading this
>>
>> https://en.wikipedia.org/wiki/Miniature_faking
>>
>> and have been wondering how to achieve the effect in POV-Ray. Obviously,
>> there is no true scale in POV-Ray, so there's no need to "fake" the
>> effect. But I don't know how to construct the correct camera.
>>
>> Any ideas?
> 
> Yep: Crank up the focal blur `aperture` setting like mad.
> 

The problems I was having involved scaling/transforming the camera. Once 
I stopped doing that, things started working properly. I don't really 
understand why, but I will post some test scenes in the other forum.


Mike

Post a reply to this message

Mike Horvath <mik### [at] gmailcom> wrote:
> I still don't understand exactly what's happening, but I removed the
> scaling and just increased the numbers directly, and things are working
> out much better.

I've never tried *scaling* a camera before, but it's an interesting idea.

Using your 'scale 2' camera, what it looks like is that the focal_blur's
focal_point of <0,0,0> (the focal 'plane')is actually shifting to <0,0,-5>. In
other words, it's keeping the same original 5-unit distance from the CAMERA--
not scaling to 2*5 = 10-units. That's what I see from a simple test, anyway.

Post a reply to this message

On 2/11/2018 10:39 PM, Kenneth wrote:
> Mike Horvath <mik### [at] gmailcom> wrote:
>> I still don't understand exactly what's happening, but I removed the
>> scaling and just increased the numbers directly, and things are working
>> out much better.
> 
> I've never tried *scaling* a camera before, but it's an interesting idea.
> 
> Using your 'scale 2' camera, what it looks like is that the focal_blur's
> focal_point of <0,0,0> (the focal 'plane')is actually shifting to <0,0,-5>. In
> other words, it's keeping the same original 5-unit distance from the CAMERA--
> not scaling to 2*5 = 10-units. That's what I see from a simple test, anyway.
> 
> 
> 
> 

I posted my scene in p.t.s-f. Could you take a look at that?

Thanks.


Mike

Post a reply to this message

"Kenneth" <kdw### [at] gmailcom> wrote:

>
> Using your 'scale 2' camera, what it looks like is that the focal_blur's
> focal_point of <0,0,0> (the focal 'plane')is actually shifting to <0,0,-5>. In
> other words, it's keeping the same original 5-unit distance from the CAMERA--
> not scaling to 2*5 = 10-units. That's what I see from a simple test, anyway.

.... and changing the focal_point to <0,0,5> does indeed bring the focal plane
back to the original <0,0,0>.

Although, for a camera position that's not just <0,0,0> but something like
<3,13,-9>, it appears that the interplay between scaled camera and 'new' focal
plane gets much more complicated.

But I tried something weird, which seems to work(!) ...

The initial simple test:
original camera position: <0,0,-5>
original focal_point: <0,0,0>

With scale 2:
new camera position: <0,0,-10>
new focal_point set to: <0,0,5>

Now, with a more complicated camera position:
original camera position: <3,13,-9>
original focal point: <0,0,0>

With scale 2:
new camera position: <...? ...> (I'm too lazy to figure it out)
new focal_point set to: <-13,-3,9>

That appears to bring the original 'focal plane' of <0,0,0> back into sharp
focus! (Although, that plane is always perpendicular or orthogonal(?) to the
camera view.) The math for the new focal_point looks to be quite simple as well.

Post a reply to this message

On 2/11/2018 11:11 PM, Kenneth wrote:
> "Kenneth" <kdw### [at] gmailcom> wrote:
> 
>>
>> Using your 'scale 2' camera, what it looks like is that the focal_blur's
>> focal_point of <0,0,0> (the focal 'plane')is actually shifting to <0,0,-5>. In
>> other words, it's keeping the same original 5-unit distance from the CAMERA--
>> not scaling to 2*5 = 10-units. That's what I see from a simple test, anyway.
> 
> .... and changing the focal_point to <0,0,5> does indeed bring the focal plane
> back to the original <0,0,0>.
> 
> Although, for a camera position that's not just <0,0,0> but something like
> <3,13,-9>, it appears that the interplay between scaled camera and 'new' focal
> plane gets much more complicated.
> 
> But I tried something weird, which seems to work(!) ...
> 
> The initial simple test:
> original camera position: <0,0,-5>
> original focal_point: <0,0,0>
> 
> With scale 2:
> new camera position: <0,0,-10>
> new focal_point set to: <0,0,5>
> 
> Now, with a more complicated camera position:
> original camera position: <3,13,-9>
> original focal point: <0,0,0>
> 
> With scale 2:
> new camera position: <...? ...> (I'm too lazy to figure it out)
> new focal_point set to: <-13,-3,9>
> 
> That appears to bring the original 'focal plane' of <0,0,0> back into sharp
> focus! (Although, that plane is always perpendicular or orthogonal(?) to the
> camera view.) The math for the new focal_point looks to be quite simple as well.
> 
> 
> 
> 
> 
> 

Thanks! That means:

     #local NewCameraFocalPoint = -1 * OldCameraLocation * 
(NewCameraScale - 1);

But is this only true if the OldCameraFocalPoint is <0,0,0>? What 
happens when the OldCameraFocalPoint is <5,20,-1>?

Would it be more intuitive if the focal point location automatically 
scaled scaled with the camera position? E.g.

     #local NewCameraFocalPoint = OldCameraFocalPoint * NewCameraScale;


Mike

Post a reply to this message

Mike Horvath <mik### [at] gmailcom> wrote:

> Thanks! That means:
>
>      #local NewCameraFocalPoint = -1 * OldCameraLocation *
> (NewCameraScale - 1);
>
> But is this only true if the OldCameraFocalPoint is <0,0,0>? What
> happens when the OldCameraFocalPoint is <5,20,-1>?

Off the top of my head, I would imagine that it's just a matter of adding or
subtracting the correct differences in values, kind of like my scale 2 example.
I haven't tested that idea though. (I'm also not totally sure if your equation
can work with an odd scaling factor of, say, 4.7 rather than 2 or 3 or 4. Maybe
it can!)

BTW: Having three active newsgroup threads, all generally about the same topic,
is kind of confusing; I'm not sure where to post follow-ups now. In future, it
would be best to have only one (or maybe an extra one for posting actual code,
if it happens to be quite long), and let the comments roll in. ;-)

Post a reply to this message

I think you need to look at it from the perspective of points, or vectors and
how they behave when scaled.

You have your camera position at a non-origin position, and then you scale it.
So it MOVES.
I'd say the same goes with the look_at point.

Try using a vector transform on a "point" and see if the scaling transform on
two vectors defined to be your camera position and look_at points give you
results that make sense.

#declare MyTransform = transform { scale { ... } }
#declare Foo2 = vtransform(Foo1,MyTransform);

from:
http://news.povray.org/povray.newusers/thread/%3Cweb.56d370be14d864ee5e7df57c0%40news.povray.org%3E/?ttop=418731&toff=5
0

Post a reply to this message

On 2/12/2018 8:34 AM, Bald Eagle wrote:
> I think you need to look at it from the perspective of points, or vectors and
> how they behave when scaled.
> 
> You have your camera position at a non-origin position, and then you scale it.
> So it MOVES.
> I'd say the same goes with the look_at point.
> 
> Try using a vector transform on a "point" and see if the scaling transform on
> two vectors defined to be your camera position and look_at points give you
> results that make sense.
> 
> #declare MyTransform = transform { scale { ... } }
> #declare Foo2 = vtransform(Foo1,MyTransform);
> 
> from:
>
http://news.povray.org/povray.newusers/thread/%3Cweb.56d370be14d864ee5e7df57c0%40news.povray.org%3E/?ttop=418731&toff=5
> 0
> 
> 


Yes, look at the code I posted in p.t.s-f. That's why I posted it.


Mike

Post a reply to this message

Le 18-02-10 à 19:47, Mike Horvath a écrit :
> What effect does scaling a camera have on focal_blur?
> 
> I have two cameras, and they should both produce the same result. But 
> they don't.
> 
> 
> 
> // focal blur camera
> camera {
>    location  <0,0,-5>
>    look_at   <0,0,0>
>    right     x*image_width/image_height
>    up y
>    aperture 1.0           // [0...N] larger is narrower depth of field 
> (blurrier)
>    blur_samples 10        // number of rays per pixel for sampling
>    focal_point <0,0,0>    // point that is in focus <X,Y,Z>
>    confidence 0.95        // [0...<1] when to move on while sampling 
> (smaller is less accurate)
>    variance 1/200         // [0...1] how precise to calculate (smaller 
> is more accurate)
>    scale 2
> }
> 
> /*
> // focal blur camera
> camera {
>    location  <0,0,-10>
>    look_at   <0,0,0>
>    right     x*image_width/image_height
>    up y
>    aperture 1.0           // [0...N] larger is narrower depth of field 
> (blurrier)
>    blur_samples 10        // number of rays per pixel for sampling
>    focal_point <0,0,0>    // point that is in focus <X,Y,Z>
>    confidence 0.95        // [0...<1] when to move on while sampling 
> (smaller is less accurate)
>    variance 1/200         // [0...1] how precise to calculate (smaller 
> is more accurate)
> }
> */

The amount of focal blur depends on the ratio between (1) the distance 
between the camera's location and the focal point location and (2) the 
actual aperture.

So, with everything else been the same, the second camera have twice the 
focal length for the same aperture, resulting in half the focal blur or 
twice the sharpness depth.

To get the same results, you need to scale the aperture by the same 
amount as the rest of the camera use for the second camera :
aperture 2.0


Alain

Post a reply to this message

Le 18-02-10 à 22:36, Mike Horvath a écrit :
> I've been reading this
> 
> https://en.wikipedia.org/wiki/Miniature_faking
> 
> and have been wondering how to achieve the effect in POV-Ray. Obviously, 
> there is no true scale in POV-Ray, so there's no need to "fake" the 
> effect. But I don't know how to construct the correct camera.
> 
> Any ideas?
> 
> 
> Mike

Use a larger aperture. Something like
vlength(Camera_Location, Focal_point)/4

If that's not enough, use a smaller divisor.

Post a reply to this message

On 2/12/2018 2:00 PM, Alain wrote:
> Le 18-02-10 à 19:47, Mike Horvath a écrit :
>> What effect does scaling a camera have on focal_blur?
>>
>> I have two cameras, and they should both produce the same result. But 
>> they don't.
>>
>>
>>
>> // focal blur camera
>> camera {
>>    location  <0,0,-5>
>>    look_at   <0,0,0>
>>    right     x*image_width/image_height
>>    up y
>>    aperture 1.0           // [0...N] larger is narrower depth of field 
>> (blurrier)
>>    blur_samples 10        // number of rays per pixel for sampling
>>    focal_point <0,0,0>    // point that is in focus <X,Y,Z>
>>    confidence 0.95        // [0...<1] when to move on while sampling 
>> (smaller is less accurate)
>>    variance 1/200         // [0...1] how precise to calculate (smaller 
>> is more accurate)
>>    scale 2
>> }
>>
>> /*
>> // focal blur camera
>> camera {
>>    location  <0,0,-10>
>>    look_at   <0,0,0>
>>    right     x*image_width/image_height
>>    up y
>>    aperture 1.0           // [0...N] larger is narrower depth of field 
>> (blurrier)
>>    blur_samples 10        // number of rays per pixel for sampling
>>    focal_point <0,0,0>    // point that is in focus <X,Y,Z>
>>    confidence 0.95        // [0...<1] when to move on while sampling 
>> (smaller is less accurate)
>>    variance 1/200         // [0...1] how precise to calculate (smaller 
>> is more accurate)
>> }
>> */
> 
> The amount of focal blur depends on the ratio between (1) the distance 
> between the camera's location and the focal point location and (2) the 
> actual aperture.
> 
> So, with everything else been the same, the second camera have twice the 
> focal length for the same aperture, resulting in half the focal blur or 
> twice the sharpness depth.
> 
> To get the same results, you need to scale the aperture by the same 
> amount as the rest of the camera use for the second camera :
> aperture 2.0
> 
> 
> Alain

You're wrong, and Kenneth already figured out why up-thread.


Mike

Post a reply to this message

On 2/12/2018 2:02 PM, Alain wrote:
> Le 18-02-10 à 22:36, Mike Horvath a écrit :
>> I've been reading this
>>
>> https://en.wikipedia.org/wiki/Miniature_faking
>>
>> and have been wondering how to achieve the effect in POV-Ray. 
>> Obviously, there is no true scale in POV-Ray, so there's no need to 
>> "fake" the effect. But I don't know how to construct the correct camera.
>>
>> Any ideas?
>>
>>
>> Mike
> 
> Use a larger aperture. Something like
> vlength(Camera_Location, Focal_point)/4
> 
> If that's not enough, use a smaller divisor.

Yeah, I got it working. But it doesn't look as good in POV-Ray as it 
does in photography. I'll post some test renders soon.


MIke

Post a reply to this message

On 02/11/2018 06:56 PM, Kenneth wrote:
> William F Pokorny <ano### [at] anonymousorg> wrote:
> 
>>
>> ..for a technique using a not yet adopted parser change which allowed
>> image_maps to be used directly in density {} blocks.
>>
> That is *definitely* an interesting and useful idea (as is your current
> workaround.) Any idea if this image_map thing is going to be included in a newer
> v3.8.x alpha or beta?
> 
> 
I've no idea. I've since closed the image_map specific pull request. As 
a request it was submitted - as suggested - in place of another adding 
image_maps and pigment_maps to density { }. I myself use the latter 
branch with both enhancements and maintaining the image_map one was 
always extra work for me as I 'git re-based' my branches over time.

In general, I've pruned back on the github pull requests I leave sitting 
in the queue though I still publish any branches I think reasonably 
solid that others building their own POV-Ray might find useful.

Having a lot of pull requests backed up on github itself becomes a 
burden to the build test machines, it slows me down on re-basing and it 
generates a lot of email noise especially for the most connected 
developers.

Aside: I am sometimes frustrated by the pace at which changes are 
adopted - while also understanding it. We must have knowledgeable gate 
keepers and planners for the code base and they have only so much time. 
Christoph, especially, has often enough saved me from my own ignorance 
or improved upon an idea and my actual implementation. We need the pinch 
point, the bottleneck for change - and to reasonably pace change given 
many have development code efforts running alongside the master branch. 


Bill P.

Post a reply to this message

Le 18-02-12 à 16:03, Mike Horvath a écrit :
> On 2/12/2018 2:02 PM, Alain wrote:
>> Le 18-02-10 à 22:36, Mike Horvath a écrit :
>>> I've been reading this
>>>
>>> https://en.wikipedia.org/wiki/Miniature_faking
>>>
>>> and have been wondering how to achieve the effect in POV-Ray. 
>>> Obviously, there is no true scale in POV-Ray, so there's no need to 
>>> "fake" the effect. But I don't know how to construct the correct camera.
>>>
>>> Any ideas?
>>>
>>>
>>> Mike
>>
>> Use a larger aperture. Something like
>> vlength(Camera_Location, Focal_point)/4
>>
>> If that's not enough, use a smaller divisor.
> 
> Yeah, I got it working. But it doesn't look as good in POV-Ray as it 
> does in photography. I'll post some test renders soon.
> 
> 
> MIke

As the ratio aperture/focal_length increase, you also need more samples. 
Otherwise, the image will get very noisy.

Post a reply to this message

Can you

a) mimic and actual camera lens with glass / ior?

b) look through a clear sheet that has a perturbation or normal that gets
progressively stronger toward the edges?

c) look though a partially transparent image map of the scene that somehow gets
rotated a little further the farther out toward the edge you get?

d) otherwise use overlapping image maps to imitate the photographic "circle of
confusion"?

Post a reply to this message

On 2/13/2018 8:54 PM, Bald Eagle wrote:
> Can you
> 
> a) mimic and actual camera lens with glass / ior?
> 
> b) look through a clear sheet that has a perturbation or normal that gets
> progressively stronger toward the edges?
> 
> c) look though a partially transparent image map of the scene that somehow gets
> rotated a little further the farther out toward the edge you get?
> 
> d) otherwise use overlapping image maps to imitate the photographic "circle of
> confusion"?
> 
> 
> 

I don't know how to do any of these things. It would probably be easier 
to just do post-processing in Photoshop.



Mike

Post a reply to this message

On 2/13/2018 11:39 PM, Mike Horvath wrote:
> On 2/13/2018 8:54 PM, Bald Eagle wrote:
>> Can you
>>
>> a) mimic and actual camera lens with glass / ior?
>>
>> b) look through a clear sheet that has a perturbation or normal that gets
>> progressively stronger toward the edges?
>>
>> c) look though a partially transparent image map of the scene that 
>> somehow gets
>> rotated a little further the farther out toward the edge you get?
>>
>> d) otherwise use overlapping image maps to imitate the photographic 
>> "circle of
>> confusion"?
>>
>>
>>
> 
> I don't know how to do any of these things. It would probably be easier 
> to just do post-processing in Photoshop.
> 
> 
> 
> Mike

I forgot about the fact that my scenes have some tall objects, and the 
Photoshop post-processing doesn't work properly in cases like this.

Will a transparent glass really make much of a difference?


Mike

Post a reply to this message

Le 18-02-13 à 20:54, Bald Eagle a écrit :
> Can you
> 
> a) mimic and actual camera lens with glass / ior?

Yes. Tou place an object shaped as the desired lense with the desired 
ior just in front of the camera. You can also add some dispersion.
This have been actually used.

> 
> b) look through a clear sheet that has a perturbation or normal that gets
> progressively stronger toward the edges?

You can use camera perturbation or use some actual transparent object 
with a perturbed normal or some actual displacement.
Then, it's just a question of modulating the pattern.
There are some samples scenes that show how it can be done.

> 
> c) look though a partially transparent image map of the scene that somehow gets
> rotated a little further the farther out toward the edge you get?
> 
> d) otherwise use overlapping image maps to imitate the photographic "circle of
> confusion"?
> 
> 
>

Post a reply to this message

"Kenneth" <kdw### [at] gmailcom> wrote:

I just noticed a dumb error I made-- some switched vector coordinates:
>
> Now, with a more complicated camera position:
> original camera position: <3,13,-9>
> original focal point: <0,0,0>
>
> With scale 2:
> new camera position: <...? ...> (I'm too lazy to figure it out)
> new focal_point set to: <-13,-3,9>
>

With scale 2, it should be...
new camera position: <...? ...> (I'm too lazy to figure it out)
new focal_point set to: <-3,-13,9> // the CORRECT vector

I hope my mistake didn't totally confuse the idea...

Post a reply to this message

On 2/13/2018 8:54 PM, Bald Eagle wrote:
> Can you
> 
> a) mimic and actual camera lens with glass / ior?
> 

I'm surprised no one has created a library of simulated cameras/lenses 
already. POV-Ray has been around for, what, two decades?


Mike

Post a reply to this message

On 21-2-2018 12:28, Mike Horvath wrote:
> On 2/13/2018 8:54 PM, Bald Eagle wrote:
>> Can you
>>
>> a) mimic and actual camera lens with glass / ior?
>>
> 
> I'm surprised no one has created a library of simulated cameras/lenses 
> already. POV-Ray has been around for, what, two decades?
> 
> 

Edouard Poor...


-- 
Thomas

Post a reply to this message

Attachments:
Download 'edouard poor - 35mm camera.zip' (51 KB)