| 摘要 |
BLOCK-BASED NOISE DETECTION AND REDUCTION METHOD WITH PIXEL LEVEL CLASSIFICATION GRANULARITY A method of detecting noise in an image (1900, 2000) of a compound document is disclosed. The image comprising pixels representing at least text data (1910,1930, 2010) and image data (1920, 2020) used to form the compound document. The noise comprises at least one of halftone, bleeding and texture noise. The method partitions (120) the image 10 into a plurality of tiles (Fig. 11) each formed of a plurality of pixels and determines (140) a set of dominant colours for each said tile and associating each pixel with one of the determined dominant colours. The method then determines (1210) for each dominant colour a fragmentation statistic (1210, Fig. 13); and applies (1220) a threshold (NoiseCnt, NoSolid, NoiseType) to each said fragmentation statistic to detect the at least one of 15 halftone, bleeding and texture noise in the corresponding dominant colour in the tile. The method may be extended to reduce detected noise in the image by processing the colours of each said tile by (a) merging each of the colours with noise to one said colour without noise if the noise is halftone noise or texture noise; and (b) removing the colours with noise if the noise is bleeding noise. The pixels of the tile are then quantised to those 20 remaining dominant colours of said set. 1064822_1 831907_speci02 |
