A binarizing unit binarizes a multilevel image. A region specifying unit A
specifies the position of a character region in the binary image, and
determines whether or not the specified character region is an inverted
character region. A region specifying unit B specifies the position of a
unit character region. A reducing unit reduces the multilevel image. A
representative color extracting unit extracts the representative color of
a unit character region on the basis of region type information
indicating whether or not a character region is an inverted character
region, position information of a unit character region, and a reduced
multilevel image. A character region padding unit pads a character region
of a reduced multilevel image with its neighboring color on the basis of
position information of the character region. A JPEG compressing unit
compresses a padded reduced multilevel image. An MMR compressing unit
compresses a partial binary image corresponding to a character region.
Compressed data containing position information, the compressed code
created by each compressing unit, and the representative color
information of a unit character region is output.