Abstract: An image processing apparatus for image compression and decompression includes a smoothing part for smoothing image data of a reconstructed image by using a smoothing filter selected from among a plurality of smoothing filters having different smoothing effects, an averaging part for producing a set of values of averages of coefficients produced through discrete cosine transformation with respect to a plurality of blocks in the reconstructed image so that a distribution of the averages of the coefficients for a target block is generated, and a selection part for generating a signal to enable the smoothing part to select the smoothing filter by detecting whether or not peak values in the distribution of the averages of the coefficients, generated by the averaging part, appear substantially on a diagonal line of the target block passing a lowest-order position through a highest-order position of the target block.

Abstract: An image area discriminating method surely distinguishes a character image area and a halftone image area printed on a single document in response to image data outputted by an image reader. An image of interest is divided into N.times.N blocks and subjected to two-dimensional orthogonal transform block by block to produce an N.times.N coefficient pattern. The blocks each is compared with coefficient patterns prepared beforehand and matching the characteristics of character images to see if the block lies in a character image area or in a halftone image area. The N.times.N coefficients may be replaced with N.times.N vectors. Then, a necessary number of vectors are selected out of the N.times.N vectors to generate a .upsilon.-dimensional coefficient vector, and it is compared with vecotrs prepared beforehand for character identification and matching the characteristics of character images.