The method decompresses blocks of a compressed image (202,204,206,208) and boundary filters the blocks (210). The method one-dimensionally filters across those boundary regions having a common boundary in accordance with a predetermined formulae, wherein the one-dimensional filtering is applied perpendicular to the common boundary.

Adjacent regions are identified in an image. Coding parameters for the adjacent regions are identified. Selective filtering is performed at the region between the identified adjacent regions.

An image processing apparatus reduces a distortion that occurs at a tile boundary (tile distortion) when an image is compressed or expanded on a tile-by-tile basis. It is equipped with a tone jump detecting section, a slope calculating section, and a slope correcting section. The tone jump detecting section calculates differences between the values of pixels on both sides of the tile boundary and detects a tone jump in the tile distortion according to the differences. The slope calculating section calculates slope-like correction values to be used for causing the tone jump to approach zero in a prescribed length ((prescribed length)<(tile length)), and adds the slope-like correction values to the values of pixels near the tile boundary. As a result, the steep tone jump due to the tile distortion is corrected into a gentle slope-like variation. Tile distortions are reduced properly by this processing.

A video frame is decomposed into blocks of pixels. Blocks that contain moving edges are not processed further. The remaining blocks are examined for the likelihood of having color bleeding artifacts, and not processed further if the likelihood is low. The likelihood is determined by examining statistics of the chrominance values. The remaining blocks are de-color bleeding filtered at the chrominance plane, without affecting the corresponding luminance plane.