y researches focus on two-dimensional artworks. Using my own software I analyse colour and shape usage in widely-known works of art. My software analyses proportions of colours in a painting (or more precisely in a digitised image) by perceptual hue, brightness, saturation and darkness (see below). I also comment for each painting on the mood conveyed by that piece of work, a conjecture has to be, for the time being, subjective.

s regards shape, this is closely related to edge detection, at the moment I do not have any software for picking out discrete shapes from an image, I think some may exist though only at an essentially elementary stage compared to what the eye and brain does all the time, making inferences based upon experience. With edge detection I use my software for reasons that are discussed at length on these pages, see for example Edge Detection Scripts.

he purpose of these analyses is to try and glean some clues about how paintings that are generally regarded as great works may be identified as such through the use of shape and colour. Much is already written about this, though I am looking at it from an angle that seeks to extract inferences that could be used in the development of automated design aids. I’m less interested in the whys and wherefores of the art works themselves.

Colour name (perceptual hue) I use twelve colour names: red, orange, yellow, yellow-green, green, cyan, blue, purple, magenta, grey, white and black. The first nine of these are defined as being within a given range of hue on the HSV colour circle (see below). Of the final three I define black as being a darkness value of less than 2%, white a saturation less than 7% and darkness value greater than 92%, and grey as either a saturation less than 7% and darkness value greater than 91% or a saturation of less than 20% and a darkness value less than 30%.

Note on the HSV colour circle: HSV is sometimes called HSB or HSL, the H and S are clear, standing for Hue and Saturation respectively. V means Value, B means Brightness, and L means Lightness, of these L, Lightness, is the most accurate, though what is the difference between Lightness and Brightness? The trouble is that Brightness means something else too, as I explain on my page Formulas for Calculating Pixel Brightness so B, or L are rather ambiguous, which is why I tend to use HSV and call the V ‘Darkness Value’, which strictly speaking is inverted as the higher the value the lighter it is, not darker, though conceptually I think makes more sense.

Note on the analysis techniques In analysing paintings I haven’t been along and chemically examined the paint, it is all from photographs. The results will therefore be greatly influenced by how accurate the colour reproduction is on the photograph. It obviously will never be 100 per cent. I think though that this doesn’t matter too much, for I am examining relative hues and brightnesses, the resulting statistic may not be rigorously accurate, but will I believe be accurate enough for what I am trying to examine.

Paintings analysed are to be found classified by artist from the submenus of menu item Art Analysis or from the list below.