Much of the lack of "dramatic perspective" when using a dome is due to using texturing images with a (nearly) linear rate of change of horizontal distance from the center, whereas they should show an exponential (squared, in particular) rate of change.  The difference will become greater as you approach the horizon.

     For those making sky dome texturing images, try using the widest angle available, so that near the edges of the image, the horizontal sky distance covered by an edge pixel (which has a greatly oblique perspective) is many times that covered by a pixel at the center (which has an orthogonal perspective).

Bookmarked - fantastic thread!

Quote - The problem with photo-based textures is that they are comprised of pixels. This is not the fault of the CG camera.

If you looked at a digital photo very closely with the human eye, quality would still erode.

Quote - Paloth,

I think you misunderstand. Kawecki is not talking about the texture, he is talking about the fact that the human eye "auto-adjusts", for lack of a better term.

Is not the problem due the texture is made of pixels or the human eye adjusts.
In renderings the pixels are stretched vertically, more near the camera more stretched and the stretching increase is very un-linear.
If you increase the number of pixels they still continue to be stretched and look as dashes.
With human eyes this never happens, if you look at a grainy material with round grains, the grains remain round no matter how near we look. The only thing that change is that grains became bigger or smaller, but always round.

Quote - Not that any of this matters, but the human eye does not change focal length. It changes the focal plane, providing a choice between near or far objects being in focus.

There's no way to change the focal plane, there are no muscles in the back of the eye. Also would be easy for a person with short vision with some muscle exercises to achieve normal vision.
If there is no way to change the focal plane and the front of the eye doesn't move to change the focal length then the only remaining possible way to change the focus is to change the refraction index of the liquid inside the eye. Perhaps this really happens.....

Ok, I have a practical question, and I hope I explain this well enough.

I read in another thread that to get the best use out of an environment that was rendered in Vue, to be used in Poser, the camera focal length in Poser should match that used in Vue.

So, let us say I wanted to render an equirectangular image in Vue of a sky and terrain to use to create an image of with an aircraft in Poser, and I render the map in Vue using a focal length of say 35mm, what size should I render that image, and should my focal length in Poser then be 25mm for the best results?

And should I use different focal lengths for different purposes. For example, if I am rendering an image in Vue to use to make an image in Poser of something small, say a render of an ant, should the focal length then be different from what I am using if I am going to render the Titanic. I am talking about the focal length in Vue that I need to use to render the image for the environment sphere.

There is no such thing as focal length when rendering an equirectangular image. Who told you that?

It's nonsense.

Use your head. Focal length combined with image size defines the field of view as we've been discussing. In an equirectangular image, the field of view is 360 degrees - an impossibility for a tradiitonal camera at any focal length.

Clearly this scenario has nothing to do with focal length.

What does matter is pixel resolution. I've written about this several times. If you're going to render at a FOV of 90 degrees, you plan to render 1/4 of the sphere. If you want that to be at resolution of 1000 pixels, then you plan that 1/4 sphere = 1000 pixels. Which means the whole sphere image horizontal dimension must be = 4000 pixels.

Suppose you plan to render 36 degrees at 1500 pixels. Then you plan that 1/10 sphere = 1500 pixels. Which means the whole sphere = 15000 pixels.

Attached Link: http://en.wikipedia.org/wiki/Perspective_distortion_(photography)

As you can clearly see, there is a problem with scale. The second image would have looked right if there was an ant the size of a bus in the the image.

So in my post above I thought this had something to do with focal length, which it doesn't, so back to the practical question, how do I know what the scale is or how to make the image so that it is in the correct scale if it has nothing to do with focal length?

If you render or photograph a plant that is 10 inches from the camera, then it will look like it is ten inches from the camera at all focal lengths.

Again, as I pointed out much earlier, perspectIve is about where you are, not what magnification you're using. A photo with the perspective of being only inches from an object will always have that perspective when mounted on an EnvSphere.

OK, that makes sense.

So, I want to render a panaramic image in Vue to use in Poser. I create a scene with sky, terrain, trees, plants, mountains, etc.

It doesn't matter what focal length I use, and I should render the image as you explained earlier when it comes to size.

So the last question is, where should the camera be. In the centre of the scene, obviously, but at ground level, eye height, or where?

It should be where you expect to see everything when you use it.

That means if you expect to look down and see the ground far away, then your camera should be far from the ground.

If you expect to look down and see the ground right under a character's shows, then your camera should be at eye height.

Whatever viewpoint you record is the only viewpoint you can play back. Remember, the magnification is separate and is the only thing that the zoom/focal length controls. It does not control perspective.
 

Thanks,

I have been rendering lots of test renders, and you are right. Images rendered with the camera at eye height (I am using 170cm) look good when rendering cars, people, etc. I am going to have to render slightly different ones for use with aircraft.

Throw me a couple of those if you don't mind. Some carefully crafted landscapes would be easier to use than most photos.

At the moment I am still rendering test images, so very small sizes like 800x400 pixels, just to test everything. Once I am happy that I know exactly what I am doing, I will render them as large as I can, and most probably make a few freebies out of them.

One of my images has 140 000 trees in it, lol, so you can see why I am not rendering at full size until I am sure it is right, it will take a while to render the large images.

Bagginsbill,

Just checking if I am right in my assumptions.

1. I make a scene in Vue, and render a panoramic image. I set my camera at eye height, 1.7 metres.

2. I load the panoramic image into Poser as a texture on the EnvSphere. I put my camera in the centre of the scene, at height zero.

3. If I load props, people, objects, I have to drop them, and the shadow plane, to -1.7 metres, to get the same viewpoint at which I rendered the panoramic image.

Is that correct? If it is, I assume that if someone releases panoramic images like this as freebies, they should then also say at what height the image was rendered, so that people using it will know how much to drop the objects?

You will find that when the sphere is sufficiently large, precise placement of the camera is not necessary. In fact, it can be out of place by quite a bit without any noticeable distortion of the perspective. If this were not the case, then environment spheres would not be nearly so useful as they are.

Technically speaking, if you desire to minimize perspective distortion then you don't depress the scene objects. You raise the center of the sphere to the height of the camera. In fact you'd make the sphere move with the camera, if the camera moves. And it doesn't matter what is in the sphere - be it sky, mountains, forest,  - as long as they are distant objects they should appear to be stationary with the camera centered.

Truthfully nobody could ever tell the difference except in an animation involving a camera that changes height by a lot.

Bagginsbill, but let's say i am rendering a panorama specifically to be used with an image of a helicopter, so in Vue my camera will be say 100 metres above the ground. When I load that panorama into Poser, with the camera at zero height in the centre, the ground will appear to be 100 metres below. So I will load my helicopter in at zero height, it appears to be 100 metres in the sky, but now if I wanted to add something on the ground, I would have to drop that object by 100 metres. Is that correct?

Yes. That is correct.

To be more specific, the actual numerical values don't matter. It is the relative positions that must all be in agreement.

In all cases, the camera should be at the center of the sphere or close to it.

The camera and sphere center can be 10,000 meters above ground 0, as long as you have the objects 100 meters below the camera at 9900 meters, it will all look the same as if the camera and sphere are at 0 and the objects are at -100.

Cool.

So, as I see it, I should make three sorts of panoramic images for exterior use.

1. Basic ground level scenes, with the camera at about eye level. The objects (such as trees) in the scene should be far enough away so that their shadows will not fall on whatever objects you are adding to the scene. So, say for instance an area of about 100 metres in diameter that is just open space (in which people can place their characters, objects, props, etc), and then objects such as trees, rocks, mountains, etc, in the distance, and various different skies.

2. A mid height scene, for use with helicopters, birds, low level aircraft, etc. The camera at about 100 metres, and a terrain that looks good from the sky.

3. A really high level scene for use with something like a 747 Jumbo jet, where the camera is above the clouds. 

Off course there are infinite possibilities, such as space scenes, etc, but i think those three will be the most useful for now.

Any other advice before i start rendering, such as what should be in the scenes, etc?
 

Those types of perspective make total sense.

Just curious - can Vue render trees and ground without a sky, and produce a transparency?

If so, it would be possible to create a layered environment, with multiple spheres. These could be stacked in interesting ways, and would allow some parallax effects when the camera moves.

Yes, I can render with a pure white sky, and save that as an alpha layer. I haven't tried it myself but i know it can. I am running a few renders now, will try that as soon as they are finished.

OK, I have an idea, but before I do it, just tell me if it is stupid or not.

The problem with tendering the quirectangular map in Vue is that you have to render a massive image to get the detail you need for a narrow field of vision in Poser.

But, most of the time, you need the main skydome for the reflections and the light, and you just need a narrow section for the actual background.

So, in Vue, when i render a panorama, I can tell it the degrees to render. So if I tell it to render from 0 to 45 degrees, I am only rendering one eighth of the image.

So the idea is something like this. Render a equirectangular image for the skydome at say a size of say 1600 x 800, to give you the reflections, etc. Then render a 45 degree section at something like 1000x4000. Then in a modeller clone a section of the environement dome (a one eighth slice) and load that just inside the dome, and apply the large texture map to that section. Use this as the background for the render.

This should give you a highly detailed 45 degree background for the image, but also correct reflections, etc, for the rest of the image.

Stupid idea? Any other way to increase the resolution without having to use a map of 8000 x4000?

Skydome is not a solution, it requires huge textures. You must find other ways.
Remember Hollywood, you only put in scene what you see.
For background you can use only a plane covering what the camera see. With animations becomes more tricky, depend on how much need to change the background with different frames you will need a bigger plane. You can tile the plane, will a good tilling nobody will notice in the moving scene that the background is repetitive (video games).
You can do more tricks. In a frame you see the landscape and a background image, then in the next frames you point the camera to see only the terrain and in next frames you see again the landscape with background, but this time the background plane was rotated and has other texture image.
If you want you can use a skyglobe for global illumination for the scene, but this time the texture is a low resolution and blurred,  you needn't detail for illumination and you do not see the skyglobe, it only illuminate the scene.

kawecki,

That is exactly what I am saying, that I use one slice of the skydome, a 45 degree section of it, for a seperate background plane to the actual shot, and the rest of the skydome is only there for reflections and global illumination, etc.

The slice of the skydome has a high resolution texture, and the rest of the skydome has a low resolution texture.

Have you experimented with "cube mapping"?

http://www.renderosity.com/mod/forumpro/showthread.php?message_id=3621275&ebot_calc_page#message_3621275

Quote - > Quote - Rightio.  But suppose I want to set my focal length to something close to the human eye, 100mm, for example?

I'll have to take some photos out of my window when there are some interesting clouds, just to show you what I mean.  I know, a picture is worth a thousand words and all that. 

The human eye is 50 mm. And Poser's focal length is wrong by a factor of 1.4. To accomplish a human eye type of field of view, you want 50 / 1.4 = 35 HAHAHAH.

Just stumbled across this. Where does the sqrt(2) come into it? I.e., what is the source of the error?

Whichway