PURPOSE: To retrieve a picture based on visual information by retrieving the picture adjusted to a condition through the use of a schematic picture where the shape, color, size and position of a constitution body are mentioned and indicating a candidate picture.

CONSTITUTION: A picture collation means 4 selects the candidate picture by collating the retrieval condition (color, position, size and shape of constitution body, relation between constitution bodies and the like) with the schematic picture stored in a schematic picture storage means 3. A schematic picture display means 5 displays the candidate pictures (plural pictures exist). A user can designate the picture to be retrieved among the pictures displayed in the schematic picture display means 5 or he can correct the retrieval condition by using the input means 1. When correction is added to the retrieval condition, collation is executed. When the picture is instructed, an output means 6 fetches the corresponding picture from a picture storage means 2 and outputs it.