Abstract: The content based image retrieval(CBIR) on colour, shape and texture is analyzed and an entropy-based image retrieval(EBIR) algorithm is proposed. Entropy is introduced to extract the features of the image. With the combination of colour property and grid descriptor(GD), and entropy eigenvalue vector descriptor is brought forward for the description of the image. Using teh entropy eigenvalue vector, we can simply retrieve the relevant images from the database. It is shown through performance experiments that this algorithm is of scale invariance and rotation invariance to some extent. Satisfactory retrieval results can also be observed for the image with analogical vision but different colour histogram.

Key words: image retrieval, unit entropy, entropy-based image retrieval, grid descriptor