PURPOSE: To perform classification, having a high correct answer ratio akin to that of a person, by performing pattern classification with a neural network.
CONSTITUTION: A print object frame of a negative film 10 is illuminated with a light source 13, and is resolved into respective points to be made photometry with a photometry part 21. A picture image feature quantity is calculated based on photometry data with a feature quantity calculating part 25. A pattern classification part 26 is composed of a neural network. A signal representing each picture image feature quantity is sent to the neural network input layer, and a classification signal is outputted from a neural network output layer. The optimum neural network for each pattern is selected according to the classification signal with an exposure quantity operation part 27 to calculate a printing exposure quantity. Printing is made by controlling a filter adjusting part 31 based on the printing exposure quantity with a controller 30.