perplexing perspective?

[clive_murray_white]

<p>Maybe someone can explain this for me - in this quick demo picture you see all the perspective lines on the right (yellow) roughly lead to a single point, as they should - but the red lines on the left, neither meet at the same point as the yellow ones nor meet at an obvious point.</p>

<p>Sigma 15-30mm @ 15mm.</p>

<p><img src="https://fbcdn-sphotos-d-a.akamaihd.net/hphotos-ak-xpa1/v/t1.0-9/10885287_881722091860507_7078091932627494146_n.jpg?oh=0df2aba057b65e519fcca54d5008648a&oe=55333D34&__gda__=1430391180_f0d528ea19e68cabb759b53e4146e0d9" alt="" width="716" height="478" /></p>

[donbright]

<p>When we look at a Photograph, because the image result is on a flat plane, either paper, or screen, we have a tendancy to think that when we accept distortion from wide lenses that the lines will be bent symetrically, or evenly right, and left. Where as we need to relaize that distortion is not just flat, but spherical, and generally is going to have a center point where all lines bend from. Keeping in mind distortion is near to far, not just right to left.</p>

[clive_murray_white]

<p>Thanks Don, I think I get what you are saying...... if I'd lined my camera up on the exact mid point of the room, (the hanging light fittings) and square with both the room divider and far wall the perspective lines would be far more likely to line up.</p>

<p>Going back to my art school days we were taught multi-point perspective as a formal geometric discipline, and in this pic, according to the theory all the horizontal lines would converge/meet at points along the horizon line/ vanishing point whilst the verticals on a single north south axis.</p>

[John Crowe]

<p>As far as this example goes you have 5 yellow lines of good reliability. The "reliability" comes from having larger subjects to project those lines from. You only have 3 red lines and they are based on small subjects in the distance reducing the reliability. There is nothing else on the left to project any perspective lenses.</p>

<p>Once you tilt the camera up and to the right, and swivel to the right and step to the left thus getting all your verticals vertical, and your horizontals horizontal, you will greatly reduce any perceived perspective distortion. It does not look like you will have much if any pincushion or barrel distortion with this particular lens. </p>

[Matt Laur]

<p>I agree with John - those red projections are only marginal as references, based on what's visible in the image. <br /><br />To help with shots like this, I like to turn on my viewfinder grid lines - can really help with squaring things up as you shoot.</p>

[jim_momary]

<p>Clive -</p>

<p>Looking at it from the perspective (pun intended) of a Monday morning engineer ...</p>

<p>On my monitor the image appears as 11.25" wide X 7.45" high. The missed intersection area of the lines can be approximated at about 0.45" X 0.20". <br>

Therefore, the error is the missed divided by the total, or about a 0.1% oopsy. (1 out of a 1000)</p>

<p>That ain't half bad for drawing lines on a photo and validating the rules of perspective by an approximation. Most would say it's darned close.</p>

<p>Jim</p>

[rodeo_joe1]

<p>Clive, you're assuming that all the windows, doors, skirting, cupboards etc. in that room have been fitted plumb and true. Unless you've checked all those projected surfaces with a plumbline and spirit-level, then those perspective lines mean next to nothing. And in any case it doesn't take much of a misalignment (maybe a pixel or two out) for those lines not to meet. And if the obvious barrel distortion isn't exactly linear across the frame.........</p>

[clive_murray_white]

<p>Intriguing Jim - there is a back story to all this, we had a bad storm 2 or 3 weeks ago which resulted in water getting into the plaster board ceiling in my office causing it to collapse, naturally I took pictures for the insurance guys. In the process of clearing up the mess I got the idea to redesign the layout of this office so I thought I'd use one of the insurance photos (@15mm) as a basis for a Photoshop/Indesign drawing, but, you guessed it, I found that the perspective lines didn't lock-in with sufficient accuracy for me to create a reasonable digital "artist's impression" of the modified space.</p>

<p>And, that instead of both left and right perspective lines meeting a single point they had the same kind of "error" as shown in the pic above.</p>

[pete_s.]

<p>You have a lot of barrel distortion in that image. Probably 5% or more.<br /> Take a look at any picture on the net of barrel distortion of a grid and you'll see how it's impossible to get proper perspective with that going on.</p>

<p>If you corrected the image for barrel distortion and then used more lines, for instance from the floor boards, and zoomed in more in the image when you aligned the perspective line I think you would get pretty darn close to the theoretical perspective.</p>

<p>PS.<br /> I don't know if there are different version of the 15-30 but look at the lens analysis below by photozone.de<br /> <br /> http://www.photozone.de/Reviews/311-sigma-af-15-30mm-f35-45-ex-dg-lab-test-report--review?start=1<br /> <img src="http://www.photozone.de/images/8Reviews/lenses/sigma_1530_3545/15mm_distortion.png" alt="" width="675" height="700" /></p>

[clive_murray_white]

<p>Thanks Pete, I think you've completely got it with that explanation</p>