A reading machine that operates in various modes includes image correction processing is described. The reading machine operates in different modes that optimize performance for specific uses of the reading machine. The reading machine receives data that specifies a mode to use for processing an image. The reading machine accesses a knowledge base to provide data to the machine for the specified mode, with the data including specific target visual elements that are expected to be encountered in processing images for the specified mode. The reading machine captures one or several images and processes the one or several images to identify one or more of the target elements in the image using information obtained from the knowledge base.