Index Cover
Image Sample

Sample Files

CUBE.BMP is a CAD generated image of a 10x10x10 cube that is helpful for visualizing how the rectifier works with both 2D and 3D objects. When you first open this file there will be no points associated with it (though this will change if you save it after performing the experiments outlined below).

As a first test establish a point at each corner of the right hand surface with the coordinates of

#         X         Y         Location
01       0         0         Lower Left Corner
02       10        0         Lower Right Corner
03       0         10        Upper Left Corner
04       10        10        Lower Left Corner
Rectification should give you the result shown below:

First try swapping the points of one side with those of the other side in the example below points #01 and #02 were exchanged as were points #03 and #04

The effect of this exchange is for the rectified image to appear as if viewed from behind.

Then try swapping the points of two of the corners diagonally opposite, in the example below points #02 and #03 were exchanged.

This result is harder to comprehend by looking at the rectified image but if you look at the source image and mentally put point #01 at the lower left, point #02 to the right and point #03 above the line between #02 & #03 you can figure out what has happened.

Even stranger results happen if you swap the points along one edge. In the example below points #03 and #04 are reversed.

This is simply a mistake. If you have more than 4 points you might be able to figure out where the problem is by examining the Residuals but other than that you'll just have to revisit your notes and measurements.

There are also a few pre-defined points files for this cube which you can load using the Alternate Points File feature. They are in the PTS sub-directory under Samples and include:


Imagine that you found a foundation (6 metres long) and a door (2 metres high) and this old photograph. Assuming these dimensions you could rectify the image and get at least an idea of it's front elevation. As we know nothing about the camera make sure the Camera distortion parameters are set to none in the Output Parameters dialogue box.

This would give you a result something like this:

From the same image you could even learn something about the roof by simply moving the points and assuming a value for the length of the roof slope. I used 2.5 metres based on the approximate proportion of the height of the corner and the length of the roof slope.

It seems to show more roof sag than you would guess looking at the original image.

If you have AutoCAD you can play with this image some more in conjunction with the drawing building.dwg


is a real world example. There should be no default point file (unless it has already been rectified and saved)

The alternate point file wall-2D.PTS is a 2D example where the points were established by survey then projected onto a 2D plane using AutoCAD.

wall-3D.PTS uses the original 3D points from the survey.

These are good samples to explore how moving a point changes things and what the residuals are trying to tell you.

If you have AutoCAD you can play with this image some more in conjunction with the drawing wall.dwg

WINDOW.JPG is a photograph of a window where you can use arbitrary 2D coordinates to get a variety of results. Try making it square but shown as though it was being viewed from the inside.


The main difficulty with using rectified photography is the problem of getting good world coordinates. It's easy if you always have a surveyor handy but it is also possible for a single photographer to get some pretty good measurements with the addition of a few simple tools.

This example uses two modified tape measures, one normal tape measure, a tripod and some sort of level (only necessary if there are no horizontal lines on the surface being photographed).

The modified tape measures can even be of the evil variety common in North America where both imperial and metric scales are present. Get a good, black, felt pen and darken the imperialist half the tape every second 10 centimetres.

These divisions are visible from quite a distance in an image with reasonable resolution and you can count them to get a dimension. Hang these tapes on the surface to be rectified, preferably from a horizontal feature, then measure the distance between them with the third tape. As long as they are hanging freely you can assume them to be vertical (and parallel) due to gravity.

If there are no suitable horizontal features, as in the example below, hang them from whatever you can find and determine the vertical relationship between the two either by noting the measurement to a horizontal feature lower down (a foundation for instance) or use a level to figure this out. In this case we used a laser pointer with a bubble level on the camera tripod.

Once you have noted the relationship between the tapes (distance apart and the difference vertically) put the camera on the tripod and take a picture of the surface and the dangling tapes. Then remove the tapes and take a second picture.

The rectified image above was created from the image below:

with points transferred from the next image using the alternate point file feature of ASRix.

The tapes were 2035 mm apart and the one on the left was 235 mm lower than the one on the right. They were assumed to be parallel because of gravity.

Up Index Cover

Inquiries to:

Generated by: CART
(Thu Sep 20 17:57:02 2007 )