Have read a few on a few forums and blogs how normal maps are superior over bump maps ? something to do with directional light?? 

High end like 3d max artists use normal, bump and specular maps! is this over kill within poser?

what are the advantages of normal maps over bump maps, in our hobbyist software if any?

Is displacement the same as normal maps ??

Would be grateful for some of the more enlightened of our community to cotribute some of their thoughts to my little conumdrum!!

Cheers Ray

Somebody else can probably give a more concise answer but here goes.

Normal maps have a very similar effect to bump maps in that they both affect the appearance on the surface without displaceing the actual mesh.   Displacement maps displace the surface of the mesh.  Displacement is better for details like wrinkles or ridges where as bump or normal maps are better for finer details like surface texture.  I dont know for sure that there really is an 'edge' for bump vs normal, to me they seem like they give pretty much the same result and perfor equally well... but I'm no expert

Quote - Somebody else can probably give a more concise answer but here goes.

Normal maps have a very similar effect to bump maps in that they both affect the appearance on the surface without displaceing the actual mesh.   Displacement maps displace the surface of the mesh.  Displacement is better for details like wrinkles or ridges where as bump or normal maps are better for finer details like surface texture.  I dont know for sure that there really is an 'edge' for bump vs normal, to me they seem like they give pretty much the same result and perfor equally well... but I'm no expert

Pretty spot on. In a nutshell, normal maps are like displacement without displacing the mesh surface. It catches lighting better with directional light than a bump map. I too am no expert but that is what I gathered from my experience working with it.

I only use displacement maps, as it is the only one that actually "displaces" and gives a result in 3D.

Every pixel on a displacement map actually become a micro-poly.

Details can be put in a displacemnt map, and this actually becomes a micro-mesh over the existing polygon mesh.

Take V4, she has some 78.000 poly's to work with.

Make 2 displacement maps, one for her face and one for her body.

That are 2 maps that you can take up to 8192x8192 pixels large each. (in Poser)

Giving you 67 108 864 micro poly's to work with for the face.

AND 67 108 864 micro poly's for the body.

I work a lot in Lo Poly.

Using a 8.000 polygon mesh, and using ONE 8192x8192 displacement map offers me the possibility to work with 67 108 864 micro polygons for detail work.

Unless you're working on a game res model, where normal maps are king, I'd do as suggested and use Normal maps for fine details like pores or weaves in clothing. Displacement is great for the larger forms and details that need to change the silhouette of a character/prop in order to look right. Distance from the viewer is another factor to consider - if something is far away from the viewer in a scene, a displacement map would be a waste of resources, making a bump or normal map the better way to go. Meanwhile, if something is close to the viewer in a scene you may need to use both a displacement map and a normal/bump map.  Ultimately, it's a judgement call on your part when to use which or if to use them in conjunction with each other.

@Teyon, I agree completey.

Distance from viewer is a very important factor.

With increasing distance, details fade out.

At a distance only a good displacement map maintains it detail level longer, as both others, bump and normal map are fake and do nothing in 3D space..

The same is true for SSS. The further away, the less the effect becomes.

Eh. That's debatable and again, depends on the situation, lighting conditions, the object in question. Lots of factors. Not saying you're wrong in anyway, just that in cases where the form isn't required to change and you're at distance - a displacment isn't needed. Now if form relies entirely on the displacement, than even at distance it's the way to go. It all depends. Good luck with your work, cliss.

Thanks Guys for your input It has helped a lot :)

The primary reason for normal mapping is to place a fairly details texture on a lores object as in gaming design.  Works perfectly for background and distant models where low poly is more advantageous in staging a scene.

Quote - The primary reason for normal mapping is to place a fairly details texture on a lores object as in gaming design.

It's also the primary reason for bump mapping, displacement mapping, and parallax mapping.

The differences do not lie in "primary reason"

Quote - Displacement is better for details like wrinkles or ridges where as bump or normal maps are better for finer details like surface texture.

Please clarify "better". Better how? Easier to make? Easier to use? Easier to parameterize? More efficient in CPU? More efficient in memory use?  More detailed appearance? More accurate shadows? More capable of vertical surfaces? Capable of overhanging surfaces? Is better here defined in terms of theoretical differences or in terms of how Poser does it? When we say displacement map, do we mean all kinds, including 3D or just the 1D kind Poser does?

BB, BB, BB, BB, BB, man, man, man, man, man,

What is fake and what is true?

Is true fake?
Or is fake true?

Displacement map is the only one that actually does something in 3D.

The others are fake and make you believe they are doing something.
Only giving the illusion of doing something.
But actually they do nothing in 3D.

You only think they are doing something, while they actually do nothing.
Well, they do something; they make you believe they are doing something in 3D.
But, they actually don’t do anything in 3D.

I hope that I am clear;
But when I am clear?

You can see right trough me, as if I where made out of pure glass.

Well, if you look straight at me, you can see trough the glass as if it was not there?
But ?
When the glass is not there?
Where am I then?

Just to say that displacement is the only true one.
Bump and his fiend are fakers.

Got it?? :-)
Or need a picture with that? :-)
Have a nice day, as I am having.

We're all clear on how a displacement map actually moves the mesh.

You guys have not been clear on the rest of the questions, as this forum has not been clear about it for years.

BUMP versus NORMAL - that is the question.

quote: "High end like 3d max artists use normal, bump and specular maps! is this over kill within poser?"

quote: "what are the advantages of normal maps over bump maps, in our hobbyist software if any"

There are advantages. I have discussed these before. Does nobody remember?

What are the performance issues of all three?

What are the ease of use issues of all three?

What are the construction methods of all three?

I have answered these before. I'm not going to keep typing those answers.

An example of what I'm talking about:

Quote: "It (normal) catches lighting better with directional light than a bump map."

ReallY? How is it better? More accurate? Or just wildly exaggerated? Can you demonstrate?

You tell me:

A) Best to worst at "catching the light" - and what you're looking for there.

B) Which is which?

Need a hint?

Render times for each by itself, in seconds:

Left 91, Middle 85, Right 93

Well, I'll bite. :D  The displacement is the one on our right. I think the normal is the one on the left and the bump is in the middle.I could be off on the las two.

Honestly though, I've always been a believer in you use what gives you the look you want/need and the rest is secondary - which is why I try not to get into "what works best" arguments too much. 

I would say normal in the middle; look at the specular highlights. Am i right?

Evidence please?

I see the opposite. I see that normal maps give the same answer without need of interpolation from a well-done mipmap. Which means that detail is retained over a greater distance. A corollary is that a normal map will produce slightly more detail given the same memory commitment, i.e. resolution of image.

Here's my evidence.

Up close.

Teyon and MG - you have to say which, in your opinion,  is doing a better job at "catching the light", whatever that means. You can't just say "based on the highlights". You have say whether and which highlights you prefer or think are more accurate, which are two different points of view that a single person can choose either.

Let me be more clear, without yet saying which is which.

To me, the middle one is missing some highlights. Is that your opinion as well? To me, that is less accurate and less desirable. But you know me - accurate = desirable for me.

Do you agree that missing some highlights is the same as saying "Not so good at catching the light"?

Normal, bump and displacement. In that order.

But it is easy to fool the eye from an exact front position.

For what it is worth?

I experimented with procedural bump and displacement and ended up "releasing bump", maintaining only displacement.

Combining the texture and a procedural displacement, and driving the displacement by the texture.
On the wrong PC right now, but I will try to post an example with explanations of what I am talking about ASAP.

Starting from upper right: the texture map, increased in size to 8192x8192, desaturated some 10% and sharpened twice.

From there I add A Math_Substract function to get a inverted map to go into  the Displacement-Blender input_1

Result? =>  Brows and public hair get most displacement.

Input_2 from this Displacement blender goes to the Turbulence.
And blending is set to 0.5 and here I can control the mix if needed..

From Alternate_Specular to the Specular-Blinn node where I reduce the Specular_Color using the Blinn-Color-Blender.

The Blinn-Color-Blender gets input_1 from the texture. => Brows, lips, public hair stay with less Blinn.
Input_2 goes to turbulence and is blended at 0.1

Specular-Blinn Reflectivity itself is driven by the texture again (output from the hvs node).

Quite a simple setup, and it gives me lots of control over the Displacement, and over the Alternate_Specular.

Comments are more then welcome.
Thanks for looking
Tony

"But it is easy to fool the eye from an exact front position. "

It's not exact front position. The boards are rotated 10 degrees from vertical and the camera is off center yOrbit = 3 degrees.

Quote - Teyon and MG - you have to say which, in your opinion,  is doing a better job at "catching the light", whatever that means. You can't just say "based on the highlights". You have say whether and which highlights you prefer or think are more accurate, which are two different points of view that a single person can choose either.

Overall they are the same at catching the light. In theory, displacement could be better, because it generates a larger surface which could catch and scatter more light, but as far as i know poser does not support raytracing with displacement, so it should be the same.

Quote - Let me be more clear, without yet saying which is which.

To me, the middle one is missing some highlights. Is that your opinion as well? To me, that is less accurate and less desirable. But you know me - accurate = desirable for me.

In the close up it does. But in the distant view i would say the outer two tiles look a bit more blurred than the middle tile, so the situation seems to be reversed, slightly.

Quote - Do you agree that missing some highlights is the same as saying "Not so good at catching the light"?

Yes, the word "missing" implies that they should be there in the first place. Missing highlights could be the result from both a smooth surface and a bad map. If a smooth surface would be the reason, missing highlights would be accurate (but not necessarily pretty). A blurry map would also miss some highlights but would otherwise be indistinguishable from a smooth surface. But (and i am taking a wild guess here about what you might have done for these examples) if you took a displacement map and converted it into a normal map, the result will always be worse and more blurred as the original map, it cannot be better than your input data. The same would be the case if it were converted the other way around. Except perhaps for the case when you convert something into a displacement map and used a too small bit depth; that would lead to quantization noise in the output and hence some more highlights. Not accurate in the common sense, but might still look better.

Quote - Starting from upper right: the texture map, increased in size to 8192x8192, desaturated some 10% and sharpened twice.

From there I add A Math_Substract function to get a inverted map to go into  the Displacement-Blender input_1

Result? =>  Brows and public hair get most displacement.

Input_2 from this Displacement blender goes to the Turbulence.
And blending is set to 0.5 and here I can control the mix if needed..

....

Now people - this is what matters. I've been showing you that the claims made about "better" results were bullshit. You guys can't even tell which is the normal map and which is the bump map.

But when I use the bump map in the displacement channel, you get something the other two can't do.

And on top of that, as Vilters is demonstrating, you can MANIPULATE a height map, whether procedural or image based, and you can mix it with other height maps on the fly. You can modulate it. You can use it to modulate color as well. Or as Vilters shows, use color to modulate height.

This is a level of control, power, and ease of use that you do not get with a normal map. These things TRUMP all the other advantages of a normal map.

And the 3D of displacement trumps bump. But the cool thing is the same height map drives either bump or displacement.

The computional savings of a normal map are nominal in this type of rendering, where the calculation of surface normal is only a small part of the rendering equation. This is very different from game graphics, where the calculation of surface normal is often half the rendering cycle.

Quote - Overall they are the same at catching the light. In theory, displacement could be better, because it generates a larger surface which could catch and scatter more light, but as far as i know poser does not support raytracing with displacement, so it should be the same.

Maybe you meant something different but on the face of it that last part is not true. Poser does support raytracing with displacement. I'm using raytracing in these image, and you're seeing the HDR sky reflected in the sample objects.

What you probably meant is that the per-pixel displacement does not get seen by raytracing rays. But full per-pixel rays ARE SENT from displaced surfaces.

So - a mirror showing these objects might make them appear different. But the objects themselves, which are reflecting other props here, are doing the right thing. Some of those highlights are from the sky via Reflection, some are of my infinite sun light via Blinn.

Both kinds of highlight appear to be missing in some places.

Quote - In the close up it does. But in the distant view i would say the outer two tiles look a bit more blurred than the middle tile, so the situation seems to be reversed, slightly.

I noted that in my other post regarding Vilters' claim that distant objects are better rendered with displacement. This is the opposite effect of the highlights, though. Missing highlights are the result of over interpolating. Extra details at distance are the result of under interpolating - the result of using a single sample as data no matter how many other samples should be included in that pixel due to distance.

It's false detail, though. It's actually aliasing.

Quote - Yes, the word "missing" implies that they should be there in the first place. Missing highlights could be the result from both a smooth surface and a bad map.

Yes it could, but my maps are great. I made them correctly and large.

Quote - But (and i am taking a wild guess here about what you might have done for these examples) if you took a displacement map and converted it into a normal map, the result will always be worse and more blurred as the original map, it cannot be better than your input data. The same would be the case if it were converted the other way around.

I did no conversion of any kind. My maps are procedural, generated simultaneously from Filter Forge, then recorded as large images - far larger than what I rendered.

Whatever effects you perceive cannot be ascribed to bad generating technique, low resolution, inappropriate conversion, or noobish incorrect application in the shader. I know what I'm doing. The differences you see are the result of the technology itself - bump versus normal.

In bump mapping, heights are interpolated and then a normal is derived.

In normal mapping, normals are interpolated directly.

This is a fundamental difference. It is at the heart of all the appearance and performance differences, small as they are.

I suppose I should now comfirm that:

Left is Bump

Middle is Normal

and, of course, Right is Displacement.

I don't disagree with you on any particular point here.

Sorry, had to crop to stay below 200K

See that the displacement comes out pretty well on the brows, lips, and, you know.

Also the skin, that has that nice uneven turbulence on it

Quote -
Now people - this is what matters. I've been showing you that the claims made about "better" results were bullshit. You guys can't even tell which is the normal map and which is the bump map.

But when I use the bump map in the displacement channel, you get something the other two can't do.

And on top of that, as Vilters is demonstrating, you can MANIPULATE a height map, whether procedural or image based, and you can mix it with other height maps on the fly. You can modulate it. You can use it to modulate color as well. Or as Vilters shows, use color to modulate height.

This is a level of control, power, and ease of use that you do not get with a normal map. These things TRUMP all the other advantages of a normal map.

Hm, i always thought that this being the strongest point in favour of normal maps; people cannot manipulate, be wrong about the scale or require you to guess the scale of normal maps. Or the offset, just count how often the question about the offset of a displacement map pops up. So it depends on what you actually want to have. If you measure the benefits of a bumpmap on how easily it can be used as a height/displacement map, a bumpmap clearly wins over a normal map. Different kind of perception, i think :-)

I cut them all on the loose; Bump map, normal map, specular map, all of them.

Only Displacement and Alternate_specular stays, driven by the texture and procedural.
The mat room settings are in some posts above, and it's effect brings more realism.

Easy to set up, and very controllable in all aspects and light conditions.
You can even plug in SSS, and render with SSS.

And memory wise? You only have to load one texture.
Ok, I used an enlarged texture in my example, but it works on normal size textures also.

I complensate Lo Poly by using a higher resolution texture.
PS, the figure above is 8.000 Poly.

Normals are widley used in games to reduce the enviroment poly count and the speed that it can be calculated. Some gaming engines crank out obsene framerates, and using normals mapping is just one of many things used to cut down on the calculations to achieve those framerates. Even thou it is one of the fastests (frame rate) way to do it, not all game engines use this technique.

All the methods have advantages and disadvantages, decieding on which ones to use depends on a lot of factors in the scene.

Some say a standard rule is to use displacement if the object is close to the camera. Bump works fine as long as it is not a lot of it. Use bump for things in the background that have a decent poly count. Use normals on anything that is very low poly. Yada, Yada, Yada....

Sticking to that theory limits what you can do with them, you can mix them as well to get results that are not achieveable using only one type.

Use what works for what you are trying to achieve.

That was a terrific explanation, bb.  This should be stickied somewhere.

simple mesh. 

This thread has gone beyond my wildest ;P Thank you to all who have taken the trouble to have contributed so far passing over their experiences and theories Thank you BB for your explanations on the subject did not know that you had already discussed this topic in another thread? Thank you BB, vilters, and Blackhearted for the renders posted giveing me a better understanding I think :/

  So if I am correct (which I doubt) there are cases where either Bump normal or displacements are useful but Bump wins over normal

Although  I must admit I was taken by the renders blackhearted  posted where the displacement in my tired old eyes just had the edge?  But that was displacement over normal and not bump !

BB I have very much enjoyed reading your posts it has helped a lot. So would I be right in saying that in your opinion Bump maps are more versatile than normal or displacement maps? with the right tweaking can achieve results that surpass the other two?

May I politely ask whether all the above wisdom in this thread is equally application to animations as well as 'single' frame images?

Poser renders animations frame by frame so yes.
You can use the above.
But the visual effect will be less.
At 25 or 30 frames a second, a moving figure becomes more or less blurred tor the human eye.

But yes, all of this can be used.

the disadvantage of normal maps is that they dont actually displace the geometry -- so normal mapped wrinkles and seams will look like displacement mapped ones from dead-on, but the silhouette of the figure/object will remain unchanged.

for example in the attached image the ridges along the front and back of the grip are done with a displacement map. you couldnt do this with a normal map because despite the ridges appearing when viewed directly, when turned to the side like this the silhouette of the grip would remain perfectly smooth. think of normal maps like bodypainting:  i can paint the appearance of a pair of jeans and a tank top on a woman, along with proper shading and shadows, but her silhouette is still going to be that of a naked woman.

the advantage of normal maps is that they render faster, they are visible in real-time in the poser viewport, and applying them is painless:  you plug in a tangent space normal map and set it to 1.0 strength and it looks exactly like it should in Poser. no math node BS, no fiddling with displacement strength values, no worrying if the end user has 'use displacement maps' checked or has messed with his min disp bounds settings, etc.

except in cases where you need to displace the silhouette of an object - such as on my USP Match - IMHO normal maps are usually the better option.

another advantage of normals that people may miss is this:
when you create an object - for example a t-shirt - and release it with a complex shader with displacement/bump/etc, often add-on texturers and tweakers will want to modify the shader for use with their own textures. for example if they want to release a texture set with some raised lettering, ribbing or fuzzy stuff. they can change, screw up or discard your displacement maps, bump maps, etc and the shirt will lose your wrinkle/stitching/seam detail. 
with a normal map plugged into the gradient bump channel, they are free to add/remove bump maps and displacement maps yet the detail from your normal map will remain unaffected.  IMO this is a cleaner way of doing things.

So I need to know if I'm wrong about another difference between bump and normal.  Since normals refer to the perceived angle of the surface perturbation, and use RGB colour space, I had believed them to be more accurate.  I'm using a displacement map in my project, and it works very well for things like insets and frames.  But if I try to make a smooth bulging window for example, I get a lot of stair-stepping.  It seems like each level of grey "plateaus" when micropoly displaced,  rather than producing a smooth shape.

I thought of using something like ShaderMap 2 or xNormal to convert the displacement to a normal map, but that wouldn't solve my problem would it?

It's not intrinsic to height maps that they stair step. It's because you use 8-bits which only offers 256 levels. People use color maps, but put the same gray level in all three channels, so its no better than an 8-bit gray map.

Use a 16-bit map gray map instead. Or a 32-bit-EXR or HDR. Then you will see no stair steps.

Normals have 24 bits per pixel. If you gave 24-bits to height maps they'd be just as good.

Thank you so much.  I'll look into my toolbox for a way to make 16-bit gray maps.  I've been using Ultimate Paint and Irfanview.  I must've forgot what year it was...

I love this type of thread! I do recall an earlier thread on the subject.

Poser accepts 16-bit TIFF files currently. You can try that.

Problem is, I don't know how to make one.  I have an 8-bit map now.  I assume just loading my texture into irfanview, blurring it, and exporting it as TIFF wouldn't work...

I don't use Photoshop.  I have Ultimate Paint and a few other free paint tools.  I also have PD Howler.  I'll look around and see what's out there. 

Quote - I'll look around and see what's out there. 

Gimp can do just sbout any file type you would need. (you may have to add plugins for some things, not sure what all comes in the main installer.)

And it is opensource.

http://www.gimp.org/