Malefactor seems to have got very 'vocal', once I said I wouldn't return. I don't appreciate being misrepresented
, so here I am
1. It is a 3d scene, and the camera's position was unknown, both in x-y-z space, and in tilt, roll, yaw..
2. A couple of those can be roughly worked out, but you CANNOT model such a situation using a ridiculous 2-variable equation.
Xtraeme can't. NEITHER CAN I, that's why I haven't been STUPID enough to try. I have taken significant time to explain that fact, but Malefactor
keeps coming on back with a CLAIM that he CAN.
So, malefactor - DO IT. Prove your point. Show the diagrams, and the 3D representation, with all the dimensions marked.
around it and say that you are almost there, or that everyone is picking on you, as you offer little tidbits that skirt around the issues - DO IT.
Finish your 'analysis' and post the whole thing.
Prove you are right.
You demanded *I* do that, but I have told you that it is IMPOSSIBLE for various reasons, and also pointed out those reasons, eg:
- the camera tilt/position
- the relative movement of close objects
- the curvature of the windscreen
On that, why don't you simply answer the key questions:
1. Does the curvature have to be taken into account in regard to the reflection?
2. How are you going to do it - do you know what that curvature is?
Now specifically as to the misrepresentations...
1. Taken apart on 3 counts (jpeg artifacts / "properly blow[ing] up ... pixels" / and on lens flares being at the "exact geometrical centre
of the image")
Firstly, can you tell us exactly where I was 'taken apart'? Tell us what you
see in the 'reflection' - does it look clear to you? Tell us
how jpeg artefacting isn't
an important factor in low contrast areas of an image.
Then, tell us WHERE in that post was I quoted as saying that all lens flares were at the "exact geometrical centre of the image"? Here's what I
ACTUALLY said to Xtraeme, referring to a specific TYPE of
reflection lens flare
[edited - woops!]- note the bolded parts:
I understand that you are merely offering a rough simulation, but it's worth noting that you can't possibly duplicate a *real* lens flare
without a full ray-tracing of the entire scene, inc. camera/lens/lens elements in question, all in 3D. That said, it's a good try, except... the
lens flare isn't aligned with the centre of the image, and the vast majority of lens flares of that type are symmetrical. That is, they are
normally found along a line drawn through the light source and the exact geometrical centre of the image. That is because they are
caused by the symmetrical, round lens elements, aperture, lens barrel, etc - it's basic optics.
What is it about the words "majority" and "of that type" that you are having trouble with??? Do NOT misrepresent me!
If you dispute what I wrote in that, pick it apart YOURSELF, and be very specific. I STAND BY IT.
2. And here Charlz claims the screen is both curved and simultaneously not, he even goes on to say, "I'm sorry, but I absolutely stand by
those comments - the point stands. It is CORRECT. " (I love paradoxes!)
Sigh. Here's what I have ACTUALLY SAID about the curvature:
you seem to have carefully cherry-picked an image showing a highly curved windscreen, and yet images like this one...
seem to suggest otherwise. Towards the middle of the screen, the curve is not great at all. Besides which, the curve is irrelevant unless we can
see what the object being reflected looks like.
Xtraeme also referred to the curvature of the windscreen - a reflection off a curve will be altered both in shape and size. When you add
that to the multiple possibilities in the angle, position and distance from the camera to the screen, there are simply too many
frankly, without knowing how curved the windscreen is, and without knowing how big, long or curved the dashboard/trim is, any wild
guesses are exactly that..
Gee, it all seems pretty consistent.. The windscreen is probably curved, but we don't know by how much. That curvature is important if we are
analysing a reflections size. Any questions? Which bit do you disagree with, malefactor?
So you see contradiction as sensical and being proven repeatedly wrong as scientific? LOL
Your laughter seems a little unwarranted. Which bits of the actual QUOTES above are wrong or contradictory? Or is just your interpretation?
And I note YOU still haven't admitted YOUR very basic error in regard to the left-right panning issue. Bit of hypocrisy there, or do you just
keep 'accidentally' ignoring it?
...your theory is faulty and was taken apart ages ago...
Gee, you better correct the site admins for putting it into the Hoax category then. And why are YOU bringing it all back up, and deliberately
misquoting, or quoting out of context?
You do realize it was Charlz who said that we had to account for the asymptotic behavior of the lens right? I accounted for that and applied
it to the scaling
??? WHERE did you do this, exactly, and how did it take into account the REST of the issues I've pointed out? You cherry pick one item, don't show
us where you used it, and ignore the rest? Highly scientific approach there.
It didn't scale as it mathematically should.
It didn't? - well, show us - simply quote what you did, and I'll explain why it didn't work. I'm always happy to help people learn where they
are doing things wrong.
When I asked for feedback, what did I get? A whole lot of BS about not accounting for "rotations."
So, let's look it over again, if that will help you. Post the exact equation, along with diagrams showing how you are applying it. (Maybe you could
save some time by thinking about them pesky "rotations" and how they *might* affect what you are trying to calculate - if you discover that the
criticisms were right, you could then apologise for wasting our time...)
The point is rotations will _not_ cause such extreme parallax.
parallax? And why did you hedge your bet by using the word "extreme"? Do you agree that less than extreme
issues might be caused? How did you calculate the actual amount?
Only _scaling_ will cause parallax.... So the only relevant variable there is the z-axis.
I'm sorry, but that is SO wrong... How about lateral position? How about angle of camera (both tilt and yaw)? How about perspective of lens (wide
angle vs tele)? (which also comes into play regarding the camera angle - the two are intertwined..) Seriously, such a blinkered approach is doomed.
Or are you misusing the term 'parallax'? Like I said, requote your equation and attempt and I'll look at it again and spell it out for you.
Please copy and paste the exact part that takes this apart.
Sure - you just post the actual calculation and images that you were using.
But I'll re-iterate my point above - you CANNOT possibly use photogrammetry in this situation, without a LOT more information.
I wait with baited breath for what I imagine will be an ultimately impotent reply
You DO realise what this sort of comment makes you sound like? I guess not.
[edit on 12-6-2010 by CHRLZ]