orthographic projection

[markraz]

Can someone help me with orthographic projection?
I'm not using ogl or dx just VS C++

So when I flatten each respective axis onto the "view cube" (left image bellow) to create 2d projections of side, top front etc....
then I draw that projection onto the 2d raster screen.

The front view is by default the z axis so if I just set all Z vectors to zero I can simply draw the flattened image on the screen as is, corect??. My other question is with the side top and rear views. After I flatten them do I "unfold" them (right image below) by rotating them 90 degrees prior to drawing them as the raster image? so would I just use a standard rotation matrices to project these onto a new vector space?

thanks you, please bear with me, I am dumb as a rock

[willmotil]

this is unclear ,
what exactly are you attempting to do ?
are you undertaking a tutorial or trying to teach yourself by attempting a task here.

orthographic projection vs perspective projection,
is about how you set up your projection matrix, its really only about that

provided you have your matrices properly organized
snaping off a image from any side should be trivial with a orthographic projection matrix,
or at least a matter of just taking and repeating the proper steps in the proper order

however you must be clear , as this is a difficult topic to discuss even with correct terms.



the view matrix can be used with orthographic projection to look at a unmoving object from positions around it
or you could rotate and reposition your object using a world matrix in front of the camera
then just get your buffer into a image when you have the proper shot of the object

i cant understand what you mean by flatten it before rendering it
doing that would probably cause artifacts most likely and mess up the rendering


though the above origin explanations are incorrect its still a good operational pic rotational origins are always zero
the correct term is probably translated object origin you have to understand that these operations are not cumulative
they can be reversed but they are done in step and order

[markraz]

this is unclear , what exactly are you attempting to do ?

orthographic projection vs perspective projection,
is about how you set up your projection matrix, its really about just that
provided you have your matrices properly organized snaping of a image from any side should be trivial

however you must be clear , as this is a difficult topic to discuss even with correct terms.


the view matrix can be used with orthographic projection to look at a object
or you could rotate and reposition your object using a world matrix in front of the camera
just get your buffer into a image when you have the proper shot of the object
i cant understand why you would need to flatten a object before rendering it

thanks for the reply. I'm not asking for perspective projection just ortho and yes you have to project it to a view cube
"projection" mean flattening in computer graphics or linear algebra for that matter. In general, the term "projection" refers to any dimension-reducing operation.

So what I'm asking for in a nutshell is the steps to take a bunch of 3D vectors/vertices
and view them on the 2d screen. Specifically the side and top views.

Once you project the various orthographic views on to the view cube, do you then rotate the view cube ? if needed?

[willmotil]

yes you have to project it to a view cube

?
wait are you trying to make a environmental reflection map
if not
it seems like your trying to do it backwards normally you already have a picture
you also map or give your vertices texture uv points that relate to the image
so each vertice has x,y,z,u,v values

you project the 3d polygon pixels built from the vertices
to a single 2d plane i.e. the screen ultimately thru the pixel depth buffer or sometimes called the z buffer
but the depth buffer only keeps the z value for proper drawing to occlude pixels

well you can do it one of two ways the second is the easiest if you made a camera class

1)
rotate the object to each of its sides in front of the camera and snap shot it
you have to have set up your view matrix or camera and the object to be in a perpendicular line
you have to make sure though that the object origin or center is really directly in line
of the camera's position and to target normal
you verify that by testing that a dot product of two normals each found by the following
A = normal (camera_origin , camera_direction)
and B = normal (camera_origin , object_center)
dotproduct(A,B) should be 0
if your cameras view normal is say pointing directly in the z direction 0,0,-1 and its position is 0,0,0
you just rotate your object to each side and position it 0,0,z somewere within your view frustrum near far plane
and snap a shot

second way
2) you can put your object in the world at some simple place like at position 0,0,0 dont rotate it at all
and just move your camera around to each side and point it at the object which of course will be at 0,0,0
then snap a shot
you usually build a camera class that defines how the view matrix behaves
like
someobject.Position = new Vector3(0,0,0);
// then move your camera
camera.Position = new Vector3(0,10,0) ; //to the right
camera.Direction = new Vector3(0,0,0); // point it here look left
// snap a shot
camera.Position = new Vector3(0,-10,0) ; //to the left well now be looking right
ect...

I'm not using ogl or dx just VS C++

are you doing this all from scratch and rendering it without dx or open gl like your learning how these work ?
well when i was younger i built my own software dx like framework just for fun i remember my number 1 problem
was in not understanding the basic concepts of how everything worked which caused me to learn
number 1 see how it has been done first instead of reinventing whats already been invented

sometimes a simple library is good to start learning with have you seen the directxtk
https://directxtk.codeplex.com/

[OReubens]

ortographic "flat" projection as you seem to be needing here is essentially about dropping one of the coordinates and displaying the other 2 coordinates after correcting them for the "unfold".

so if you have a cube of sides with length M, then the projection of a point (x,y,z) assuming unfolding as in 1

            top
back  left  front  right    (<- note that all these projections will preserve y)
            bottom
             ^ (and all these projections preserve x)

none of the projections can really 'preserve' z, since that "pops out" or "puts a hole into" your screen.

front: -> (x,y)
left: -> (-z,y)
back: -> (-M-x,y)
right: -> (M+z,y)
bottom:-> (x,-z)
top: -> (x,M+z)

this assumes the front has the 0,0 in the left bottom corner, you may need to shift the entire image by (2M,M) if you need it to fit on visible coordinates if your visual projection starts at 0,0.