Drug tablets morphology is important in illicit drug investigation. The tablet shapes and impressions can be used to classify drugs at crime scene. However, it is a difficult task for law enforcement officers to correctly classify drug tablets on site through naked-eye examination without utilizing proper drug tablet databases. To better assist in-vestigators in drug classification, we have developed an online drug tablet image retrieval system based on content-based image retrieval techniques. The proposed system provides a practical and rapid tool for investigators to classify tablets by morphological charac-teristics. The correct retrieval results depend on the precision of keywords. However, it is also difficult for each user to describe an object using the same exact keywords. The proposed system uses features, extracted from the content of an image directly, as query data instead of human-described keywords. It can avoid the uncertainty caused from sub-jective keywords. From experimental results, the proposed system shows its capability of the drug tablet image retrieval.
