The results of the segmentation will be determined by the resolution of the points used to map the colour space, as selected in the "resolution" field. This specifies the distance between the points in terms of the standard deviation of the noise on the data. Its purpose is to blur the space according to the estimated noise, in order to avoid identifying noise fluctuations as statistically significant peaks. Entering 3 as the resolution (i.e. 3 standard deviations) will ensure that all statistically different regions are identified in most natural images. However, this will also identify regions such as coloured shadows. Whilst this is statistically the correct result, most people prefer a result which is slightly under-segmented i.e. small regions have been merged back into larger regions of similar colour, to avoid identifying coloured shadows etc. Therefore, the resolution can be raised to 5 or higher to eliminate these small regions. Raising the resolution will also significantly reduce the computational time required.
In order to perform a colour segmentation, ensure that an RGB image has been loaded. Then select the resolution as described above. Finally, press the "Segment JK" button. The resulting labelled image will be pushed to the top of the stack, and can be displayed in the Imcalc Tv. Binary images of individual regions can then be generated using the label selection functionality described below, and may be used for masking the original image. A mean colour version of the labelled image can be output using the "Mean Col Output" button.