Colour histogram equalisation of multichannel images

Histogram equalisation works well on single-channel images; however, the technique used is ineffective on multiple channels because there are an infinite number of ways to perform the equalisation. A technique is developed that works on images with any number of channels, and the inadequacy of a direct extension of histogram equalisation to multiple dimensions is shown. A logarithmic metric is developed to express how well the available histogram volume is utilised and how well equalisation is working. Results of using the technique and using it in reverse, i.e. to reduce the histogram utilisation, are shown. Extensions of the technique are used to measure the difference between and conform one multichannel histogram to another, and uses of these techniques are discussed.