Colour object location using complex coding in the frequency domain

Much research has been conducted on recognising objects in monochrome images using frequency domain processing. This, however, ignores the useful information that can be contained in colour representations, assuming that there is a colour camera available. The authors use a technique which is believed to be novel to represent the chromaticity information in an image. They then develop the principles of monochrome processing further by using this representation to assist in object location. In the case of Fourier transforms, a significant advantage of the colour representation is that to transform the chromatic image into the frequency domain requires only one Fourier transform whereas using a model requires three FFTs to be calculated. Two techniques for locating objects, phase correlation, and the cepstrum procedure, are described and the advantages of combining the colour representation and location techniques are illustrated. They show that, using this colour representation, both methods are able to produce an intensity peak corresponding to the displacement of an object in comparison with a reference image. The advantage of using this colour representation is that the colour of the displaced object is calculated as part of the location procedure and does not require any extra processing. If the intensity image were used in the location procedure, the object would be found but it would not be possible to estimate its colour