TEXT REGION SUBTRACTION AND TEXT RECOGNITION METHOD OF COMPRESSED NEWS IMAGES

摘要

PURPOSE: A text region subtraction and a text recognition method of compressed news images is provided to determine whether text exists in a frame by using a DCT(discrete fourier transform) coefficients rather than decoding news images in order to reduce time required. CONSTITUTION: The text region subtraction and a text recognition method of a compressed news images includes following steps. When the number of DCT blocks whose horizontal, vertical and diagonal coefficients from an I-frame configuring compressed news images are more than predetermined values is over a predetermined value, than the I-frame is determined as a text frame(11). The text region at which a gradient between pixels from the determined text frame is more than a predetermined value is determined(13). Then, the text region is converted to a binary format(14). The determined text region is divided into a text and background region, and the background region is masked. The output of the preceding step is filtered so as to reduce noise(15). The binary text region is divided into separate characters(16) and the characters are normalized(17). The normalized characters are recognized(18).