Characteristics of image data from a reading unit are unified for storage. The state of the image data is recognized and retained as auxiliary information. The auxiliary information is checked, and an image skew and the like are detected according to a target output format to perform a correction process. At that time, an optimum image processing path is selected for processing according to the state of implementation of hardware and software. A simple process can be performed at an image processing apparatus, while a complex process is performed at a network-connected PC. Since a unit configuration can be selected according to the purpose of image processing, processing can be performed with a simple configuration at high speed.