Establishing VPs and horizons in Landscape photos?


Hey, no worries, there’s no hurry. :slight_smile:


Here we go:


Slightly sharper?

For larger, go to this link:


A couple of things I didnt mention before. The grid that I drew out for the ground plane consists of perfect 1m X 1m squares. Its also easy to measure exactly what distance any point in the scene is from the camera.
One other thing I need to find out about regarding field of view is that i know in this image I have plotted a 60 degree cone of vision and translated that into a HD aspect ratio framing of the same field of view, but I have only seen the field of view constructed from where the 60 degree field of view lines intersect the horizon. This makes for pretty uninteresting framing, I need to do some research into how the FOV changes when the camera is pointing up or down. I would imagine it should stick to the lines and then at the horizon the lines should converge again at the same angle above ghe horizon? I’m not entirely sure, perhaps someone else here has the answer?
If the perspective setup here isn’t clear, tell me and i’ll try and explain more clearly :slight_smile:



Sorry for the delayed response - I’ve been away for a few days. I’m off on holiday again on Wednesday - not sure if I’ll have time to experiement with this before I go but I’ll be in touch shortly after I get back. I’ll see if I can work out how to adjust the horizon line or see if I know anyone that can help.

Much appreciated for the info, thanks a lot.



I just looked in my book and this is actually covered very briefly, I didn’t know what I was looking for before but ‘perspective Cone of vision’ has thrown up a few diagrams on google too.

I’ve not got time before my holiday to get my head around it properly, but I think this is what we’re looking for.

I’ll CAD up some diagrams soon and post here, sorry the res is pretty low.


I’m not sure if this is correct. The bright yellow lines indicate the tilt of the camera, in this case 8 degrees measured against the horizon line.

One yellow line is measured 90 degrees to the other at E2 on the RHS and should give another vanishing point - (but does it really>?!) what that does, I don’t know! It can’t be verticals as they’d be going the wrong way which makes me wonder whether this is completely wrong.

So this may (or may not) demonstrate how to draw a white square in the correct perspective using 45 degree VPs and ignoring the converging lines that exist in the photo (as that would defeat the purpose of the excersise)

So I’ve tilted the ground rather than move the camera… I think…

I guess next is - how to get your vertical VPs.

Blue = FOV 58degrees
REDs = 45VP, the faded one level with the COV measured 45degrees, and the darker one at the horizon, squished a bit.

The other lines are to measure along the arbitary ruler at the bottom, so the closest corner of the square begins 2 units right and 1 back. helped by the 45measuring points.

Is it correct…?


So the above diagram is correct apart from the yellow lines. Can’t find the orig PSD to edit.

Anyway, move the yellow lines up to the next horizontal (faded) green line, and flip vertically, this will give you your vertical vanishing point. I tried it on a new sketch and it worked - well at least it looks right!

And this helps too:


Hey, sorry it’s been a while! The diagram looks good, interesting method there to get the vertical vanishing point. I’ll have to deconstruct a 3D scene at some point so I can do some exhaustive testing to see exactly how to get a specific camera FOV, but the diagram you found offers some good insight, thanks for that. Look forward to testing all this stuff out when I get a chance.



Your diagram was correct to use the camera’s FOV to establish the CP, and from there the rest of the diagram can be drawn.

And to get the FOV, you’d need some converging lines and then construct the diagram in a different order from that.


deleted posted


To be more specific - here is my grid setup so far. The pink and green blobs are VPs that are 45 degrees to one another however not 45 to the median line from the CP.

What I want to know is how do I rotate these accurately around the CP as I want to add buildings that are rotated 30 degrees (in plan) to this grid.

Note: I used the 3rd (vertical) vanishing point to establish the horizon line using the yellow lines - 90 degrees at the RHS (or LHS) 45VP. And it seems to work so it would be useful to get the horizon line if the buidings were cylindrical for example.

thin red = FOV
thin green = 45VP
thick green = discovering what I thought I knew about perspective is not correct! :slight_smile:


I think this is correct! Not tested it on another image yet.

So the aim in this case was to establish where my Horizon line is and also where my RHS VP should be in relation to the LHS (orange blob) which I’ve picked as my LHS VP.

  1. Use Camera FOV to stablish Control point (Red)
  2. Establish 45 VPs (Dark Green)
  3. Use vertical converging Cyan lines to establish vert VP. (Note: I could tell my image was not straight so I rotated it so cyan lines were central)
  4. Use yellow lines from vert VPs and either of the 45VPs at 90deg to establish horizon line (or could have used second set of converging cyan lines)
  5. Draw an arc (or circle shown in Blue) to establish Auxiliary Control point - and it’s from here that you can specify any angular rotations of objects in your scene.

Bright Green lines = VPs based part of the RHS building

The LHS orange blob was chosen as LH VP, and the RHS could be constructed from this…

Hopefully I won’t find another problem with this setup!


Any chance you could send me that image? I’d like to try and work a few things out myself when i get a spare moment, as you seem to do a few things different to how i expected and i’d love to see whether i get the same results :slight_smile:


Sure no problem, what’s your email?

edit - attached a tidier image which I knocked up in CAD, and you can see that my black linework (establishing converging lines from the photo) doesn’t line up on the horizon line perfectly - I wonder if this is due to lens imperfections or just error on my part? Perhaps I didn’t get the rotation perfect - either way, it should be good enough for my matte painting.


Ok, after seeing your new perspective mock up, i’d say don’t send me the image. You confirmed my suspicion that the lines in the image may not vanish to the horizon. However, I think you also landed on the reason behind that - lens distortion. You’ll never get a perfect lineup with an image that has lens distortion. So really we need to test an image taken by ourselves, where we have all the lens info and can undistort the image before setting up the perspective…
I’m impressed that you’ve taken this so far already. Can you explain what the blue circle is representing? It’s possibly just a different way of working, but i’m not sure what that is being used for.
Great stuff tho, looks like you cracked the vertical vanishing point at least!


The blue circle is the method to establish where your Auxilary CP is from your (Pink) horizon line (well at least I’m prtty sure that that is the case), I drew a smaller blue circle just to make it clearer where it’s centre is.

And the radius of the big blue circle is based on the yellow lines.

So now you could draw 45VPs from your Aux CP to where the blue circle intersects the pink horizon line.

I think for sketches and paintings, the error (perhaps because of distortion) is acceptable. I think for a future cityscape full on matte painting - 3D all the way from me however I’ve yet to actually test this by setting up a 3D scene. I will do though - just not at the moment.

I’ve found that there are a few different methods to establish VPs, and I’m prety sure that this method is more accurate than at least one of the other methods - and I have to say again: that is on the provision that this is correct of course!

I’ve derived this information from a book and this web page (search for Exact Horizon Line Method:


What you have looks pretty good to me. The only real test will be with a 3d scene of course, but even a 3d scene i believe is only accurate up to a point. Like you said, the result you have there is very close, close enough to get away with it. And yes, 3D will always be quicker than going crazy and drawing all this stuff out manually, but it is great practice and training, I’ve learned a lot from all the questions we’ve been throwing back and forth in this thread.
By the way, that web page is gold! Great find. I’ll need a full weekend just to read it through without getting confused i’m sure, but its a great resource!



Yeah for sure I’ve learnt a lot too - thanks very much for your input. That website is very detailed - I just scanned over the top of it until I stumbled across a diagram that looked similar to the method used in my book.

I think this will be a handy bit of knowledge to have - I’m going to have a breather from perspective learning though - on to the next challenge for now.


This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.