Content-based image retrieval is an important problem in computer vision. It consists of finding a set of images that present content that is similar to a given query image. Since we have learned that histograms constitute an effective way to characterize an image's content, it makes sense to think that they can be used to solve the content-based retrieval problem.
The key here is to be able to measure the similarity between two images by simply comparing their histograms. A measurement function that will estimate how different, or how similar, two histograms are will need to be defined. Various such measures have been proposed in the past, and OpenCV proposes a few of them in its implementation of the cv::compareHist function.
