Distance measure between colour histograms

HenrySmith

Hi all,

I am implementing a simple Content based image retrieval. I performed the feature extraction from the images successfully. These includes colours, shapes and textures.

Now, I would like to perform similarity measure between HSB color histograms (quantised in a 7x3x3 for a total of 63 bins). For this I was thinking of first calculating the Euclidean distance in a bin-to-bin approach and then use Manhattan distance between bins to come up with a single value that represents the distance between the two histograms.

However, from some research perform I found that algorithms such as Quadratic-form distance utilizes information across bins to retrieve more perceptually desirable results.

What do you think is the best approach to calculate distance between two colour histograms (or even other histograms in general)?

I would also like to ask if there is any external java library or contributed library that contain various distance/similarity measure algorithms?

One last thing...is an HSB color histogram considered as a 3D histogram because of the Hue, Saturation and Brightness, or a 1D histogram?

Thanks