||Image segmentation, one of the most important steps leading to the analysis of processed image data, could be defined as clustering or grouping in spatial domain of the image. It often combines several higher level processing techniques, and begins with pixel classification. There are two major families of methods for pixel classification in colour and multispectral images. The more common ones are multivariate classifiers, e.g., Bayes classifiers, applied directly in the RGB colour space. With sufficient a priori knowledge, they may provide satisfactory results in many applications, but without consideration of the physical properties of light. Another group of methods, colour space transformations, is originally developed to correspond to human perception of colour. As colour information is decoupled to brightness and chroma components, a typical colour based classification is equivalent to slicing the transformed colour space with half-planes emanating from the achromatic axis. In this seminar I will describe a novel approach inspired by the method previously developed for fluorescence microscopy and then discuss limitations and possible ways of how the following segmentation steps may look like.