Perspective transformation

This function applies a perspective transformation to the current image.

A perspective transformation projects 3D points onto a 2D plane, which is used to simulate the relationship between an object and its image shot with a camera. A perspective transformation can be represented with a 3*4 transformation matrix M. The relationship between M and the camera parameters can be modeled as

Given M and a 3D point with coordinate (X,Y,Z), its imaged coordinate (x,y) can be computed as:

In the following dialog, you can specify these camera parameters and VIPBase will compute the projection matrix M or you can provide M yourself.

Note that if you used the camera calibration tool of VIPBase previously, the computed camera parameters will be automatically loaded here.

You can test the perspective transformation by providing points with known 3D and 2D coordinates. After you click the "Project" button, the groundtruth 2D points and the projected 2D points are marked in red and blue, respectively. One example is shown here.

You can visually test the perspective transformation by generating images from virtual cameras. First, click the "Select region to reconstruct" button and select a region on the image. Then click any of the three buttons below to generate image from a virtual overhead camera in x, y or z direction. If you skip the step of "Select region to reconstruct", VIPBase will consider the whole source image. Be aware that this may result in very large images. VIPBase sets an upper limit of 3600 pixels for the larger of the width and the height. Examples of reconstructed images are shown here.

VIPBase also supports the Affine transformation, another widely used geometric transformation. See the documentation here.

Last update: 08/01/2006 (check html)