This invention relates to a character segmentation method for use with an optical character reader for sequentially reading postal code, state name and city name postal information, and for generating signals which represent character masks of each character. The system then calculates the gap between each mask and determines if it is greater than an average gap value calculated for all the masks. A determination is also made if a substantially smaller mask (representing a punctuation mark) exists between the two masks. These determinations are used to identify and separate the blocks of information. Once the blocks are identified, the block containing the postal code is classified since it necessarily contains five characters and the two remaining blocks (i.e., state and city) are classified since they, having been scanned in sequence, follow the postal code. Once classified, this information is used to facilitate the character recognition process since standard patterns of an alphabetical nature can be selected when the state name and city name are being recognized, and standard patterns of a numerical nature when the postal code is being recognized.
A method of cursive address recognition of mail pieces using adaptive dictionary management is provided which includes creating a plurality of word databases containing the most frequently observed words in a particular portion of an address block derived from training data. Word images from various portions of the address block are extracted and the word images are compared to the database corresponding to the appropriate portion of the address block using a cursive word recognition engine. The search order of the word databases are updated based on the frequency of occurrence of recognized words in a predetermined number of previously evaluated addresses.
A polyphenylene sulfide resin composition comprising (1) a polyphenylene sulfide resin having a melt flow value of at least 1,000 g/min., (2) silicic acid and/or a silicate, and (3) a silane compound composed of a vinylsilane and an aminosilane and/or an epoxysilane.
Each row of characters is extracted from document data representing rows of characters by a character row extracting circuit. The characters are separated from the row by a character separating circuit. In a reference line detecting circuit two histograms are formed, one pertaining to the upper sides of the rectangles bordering the extremities of the characters, and the other pertaining to the lower sides of these rectangles. Two reference lines are defined from the histogram. The characters are sorted intoseveral categories, in accordance with their sizes and the positions they take with respect to the reference lines. The pattern of each character is compared with the dictionary patterns of the same category.
According to a method of the present invention, an optical alphanumeric character reader scans each printed line in the direction of its height from its beginning to its end sequentially at a predetermined interval. The intervals thus scanned are identified either as containing a printed part or as being all white and are represented respectively by a "black bit" and a "white bit" corresponding to different logical numbers. For each white bit or each continuous array of white bits sandwiched between black bits, the number of white bit or bits between the sandwiching black bits is counted. The histogram of such counted numbers has two peaks corresponding to gaps between letters and those between words and serves to determine a threshold value by which a space between words can be detected by the optical character reader.
A character region extracting method adapted to character recognition comprises the following three steps. The first step is to segmentate a first region (BL1) from a document image by extracting projection of the document image in a first direction parallel to a character string and detecting a range in which the projection is continuous in a second direction perpendicular to the first direction. The range defines the first region. The second step is to segmentate a second region (sBL1) from the first region by extracting projection of the first region in the second direction and detecting a range in which the projection is continuous in the first direction. The range defines the second region. The third step is to segmentate a third region (BL2) from the second region by extracting projection of the second region in the first direction and detecting a range in which the projection is continuous in the second direction. The range defines the third region. The third region may be joined together so as to form a fourth region containing characters or a character string.
