Color coherence vector

The color coherence vector (Engl. Color Coherence Vector , CCV ) is a histogram -based method for determining the similarity between two images.

Each color value has a tuple , indicating the number of coherent and incoherent pixels in the image. Coherent means that the respective pixel is part of a coherent, similarly colored area. ${\ displaystyle (a, b)}$ ${\ displaystyle a}$ ${\ displaystyle b}$

The coherence vector then results from the totality of the tuples of all colors:

${\ displaystyle \ mathrm {CCV} = \ {(a_ {1}, b_ {1}), ...., (a_ {n}, b_ {n}) \}}$

The disadvantage of pure color histograms is the lack of spatial information. Images that appear differently can have very similar color histograms, for example an image with a checkerboard pattern has the same histogram as one in which half of the image is black and the other half is white.

Color coherence vectors do not know the exact positions of the objects in the image area, but they can distinguish whether a certain color occurs in a few large areas or in many small areas. In contrast to pure color histograms, this method enables a finer distinction between images.

Individual evidence

^ Comparing Images Using Color Coherence Vectors . PDF ; 581 kB. Retrieved February 8, 2014.

[1] Comparing Images Using Color Coherence Vectors . PDF ; 581 kB. Retrieved February 8, 2014.