2009 Correction filter to reduce the fattening effect on stereoscopic images
A new method developed by reserchers of the University of Balearic Islands and the Centre National d'Etudes Spatiales for correcting the geometrical disparities in 3D images.
Geometric method to correct disparities in 3D images, improving their accuracy regarding reality through the implementation of a new algorithm in the phase of image processing.
A new method for correcting the geometrical disparities in 3D images improves their accuracy by implementing an algorithm in the image processing phase.
Taking a pair of stereo images as a starting point, which has been obtained from the classical sum of squared difference -SSD- block matching method, the designed algorithm improves its accuracy reducing the well known fattening effect, which occurs especially along the edges of buildings and tiles due to the existence of geometrical disparities caused by the presence of different depths.
The adaptive algorithm consists on using a weighted SSD. The new-discovered weights have been mathematically proven to be optimal. The resulting computed disparity is mathematically proven to be the convolution of the real one with a fixed low pass filter. This means that with this new method the obtained disparity is well sampled, contrarily to all previous methods.
Some experiments have been taken on satellite images in cooperation with the Centre National d'Etudes Spatiales.
But the method could be applicable on other sectors such as medical imaging, photography, robotic vision or electron microscopy.
It might be of interest for companies involved in the digital image processing and particularly in the stereovision field. Therefore, the interest is focused in the search of a company interested in developing the commercial opportunity.