Results

The results are very satisfactory.

These next pictures are an example of two rectified images produced by the system.

As you can see, there is really nothing to see this way. In this project I used another way to show pairs of images. It's based on a bi-color image, on which each image fills a different color, red and blue. The next image is the same pair of images, using this new method to represent them.

As we can see, this new method allows us to compare and evaluate the images more easily.

In the previous image is also indicated the measured precision of the implemented algorithms. A maximum discrepancy of 2 pixels was obtained.

This image was obtained in another test of the system. This time the calculated matrix was more precise, leading to 1 pixel maximum discrepancy in far-away zones.

Although the cameras are visually rectified (see Pictures), the images without rectification may have more than 10 pixels of error. For example, the next image represent a pair of images before rectification, where we can see a vertical discrepancy (error) of up to 17 pixels.

The images presented show that the implemented algorithms are capable of rectifying two videos in real-time, and with a good precision.