Chromaticity Analysis

This webpage allows you to use an image-processing algorithm to transform an image from the traditional Red-Green-Blue colour space to the Hue-Saturation-Value colour space. The HSV model describes perceptual colour relationships related to the artistic ideas of hue, tint and shade. While hue is a circular property related to the wavelength of the colour where red corresponded approximately to 0o, yellow to 60o, green to 120o, cyan to 180o, blue to 240o and magenta to 315o. Saturation indicates the purity of a colour (saturation = 1) or how close to white or grey (saturation = 0) and thus devoid of colour. Value is related with the darkness or brightness of the pixels.

For this algorith, hue, saturation and value have been quantised to 32 levels. The output of the algorithm is a 3D HSV histogram ( mHSV(h,s,v) as explained in the reference below) as 32x32x32 matrix, where the value of each point in the matrix is the number of pixels with the corresponding (h,s,v) values. The 50% ratios (the amount of pixels that rest in one half of the matrix relative to the total, for each dimension) for hue, saturation and value are also provided. The code to plot the 2D and 3D histograms will soon be added to this webpage.

The Chromatic Analysis is used as part of a segmentation algorithm for cells stained through immunohistochemistry. The following article describes the algorithm, the transformation from RGB to HSV and the 1D/2D/3D chromatic histograms in detail: C.C. Reyes-Aldasoro, et. al., An automatic algorithm for the segmentation and morphological analysis of microvessels in immunostained histological tumour sections, Journal of Microscopy.

You can upload an image of maximum size 2 Mb of the following formats (ONLY) TIFF, BMP, GIF, JPEG, or PNG. The name of the file must ONLY contain letters uppercase and lowercase (a-z, A-Z) and underscore (_). Any other characters such as -, +, %, ^ and blank spaces may create problems. A single dot can be used to distinguish the name of the file, from the extension (e.g. Image24_a.jpg is acceptable).
Press Send file to continue.


Input image:

Segmentation of Endothelial Cells from Images Stained through Immunohistochemistry
The University of Sheffield, Western Bank, Sheffield, S10 2TN, UK