Shape Based Image Retrieval: A Review

Content Based Image Retrieval (CBIR) system is an approach to search images and retrieve relevant images from image databases using visual content (information) of an image. In today’s scenario, CBIR receives a query object as input and retrieves similar objects as output from an image database according to a similarity/distance measure. Generally, color, texture, shapes or any combinations of them could be used as visual contents (feature) for image retrieval. Among them, shape is one of the most important features used in CBIR. While performing Shape based retrieval, it involves three primary issues in retrieving images namely Shape Representation, Shape Matching (Similarity Measures) and Shape Indexing. Also, it is important that the descriptors need to extract the characteristics of objects being robust to image rotation, scale changes, illumination variation, occlusion, noises and change of views etc. Motivated by the above factors, this paper gives a short review involving Geometrical and Structural Shape Representation of descriptors and a few Shape Matching algorithms. Keywords-shape descriptors; similarity measures; geometric matching; structural matching