Abstract: This study presents an indexing and a quick sort method that could be used to find the possible matching guilty pictures from the police database by using a tracing software and eyewitness descriptions about the suspected guilty profile. It is well known that databases located in the police departments have a large amount of prerecorded convicted guilty pictures. Searching a picture from database for recognition is very hard and it may take too much time. In addition to this, it can also require to track many redundant pictures to find the correct guilty. There is, of course, no guarantee that Photofit (robot picture) is included in database. Therefore police departments have difficult problems to recognize exact guilty using Photofits described by eyewitness. In this paper, a matrix construction using particular distances on face has been demonstrated. In order to find these certain points on the face, a new point selection schema and some existing recognition methods are presented. Finally, three different determinant values of this matrix are computed and are used for picture indexing in the database. In the test environment, the system results 30 different matches with the predefined Photofits by using this indexing algorithm together with a face tracing software and a patch program, proving that the tracking time for redundant pictures is reduced.