Abstract: A data compression apparatus is capable of compressing data at a high compression rate and at a high speed. A data decompression apparatus is used in combination with this data compression apparatus to decode the compressed data. The data compression apparatus includes a RAM which stores Huffman code trees each of which corresponds to a character string, i.e. "context." A CPU encodes each character which is to be encoded by use of a Huffman code tree corresponding to the context at that time. Each time a character is encoded, the Huffman code tree used is reconstructed so as to account for the encoded character. The data decompression apparatus stores Huffman code trees corresponding to contexts respectively, decodes the code to be decoded by use of a Huffman code tree corresponding to the context (a character string previously decoded). Each time one character is decoded, the Huffman code tree used for decoding is reconstructed so as to account for the decoded character.

Abstract: A data coding system is arranged such that character font data is compressed in a unit of 4-bit in a form of a bit string with a compression width of 4-bit in the Huffman coding system. The data coding system employs a code structure in the Huffman coding system for respectively converting three bit strings selected in order from that of the highest appearance frequency in a unit of 4-bit in the character font data into the compressed codes expressed using not more than 3 bits. The described system permits a significantly improved compression rate of the character font data into respective compressed codes.

Abstract: This paper examines the use of arithmetic coding in conjunction with the error resilient entropy code (EREC). The constraints on the coding model are discussed and simulation results are presented and compared to those obtained using Huffman coding. These results show that without the EREC, arithmetic coding is less resilient than Huffman coding, while with the EREC both systems yield comparable results.

Abstract: An amount of data coded in the past is detected at a prescribed period. Criteria for deciding whether a pixel block of interest is to be coded by processing for motion-compensated coding of interframe difference, processing for coding of interframe difference without motion compensation or processing for intraframe coding are adjusted based upon the amount of code detected.

Abstract: PROBLEM TO BE SOLVED: To provide the image coding data generating method in which image data are scrambled without deteriorating the image quality of an original image. SOLUTION: In the image coding data generating method, image data are divided into blocks each of number of longitudinal and lateral pixels is a prescribed number, discrete cosine transformation(DCT) is applied to each divided block, data subject to DCT are quantized and the quantized data are converted into a quantization DCT coefficient block group so as to apply modulation coding to the image data. In this case, a Huffman table used to apply modulation coding to the quantization DCT block is generated from a quantization DCT block group and the generated Huffman table is used, a DCT block coefficient is subject to Huffman coding to generate image coding data, the Huffman table is encrypted and the encrypted Huffman table is inserted to the image coding data to obtain the image coding data of the image data.

Abstract: A new chromosome encoding method named homogeneous coding is proposed for genetic algorithm based optimization of numerical parameters. Homogeneous coding generates a bit string code in which each bit contributes equally to the value represented by the code. Homogeneous coding and its generation method are introduced, and five basic characteristic are proposed to compare and analyze chromosome encoding methods. By using these five characteristics, homogeneous coding is compared with conventional coding methods such as binary coding and gray coding. It is proved experimentally that diversity of the gene pool is maintained in homogeneous coding better than in conventional methods and homogeneous coding is effective for optimization of numerical parameters whose searching space is complicated.

Abstract: The PPM algorithm is known to be one of the most powerful data compression techniques. However, it uses a large amount of memory, and operates slowly. Several new techniques were explored which address the problems with PPM, at each order of context modeling: the DefSum (or deferred summation) order-0 coder, the ListLRU order-1 coder, the static Huffman order-1 coder, and the PPMCB high-order coder.

Abstract: The authors propose an efficient memory allocation method for memory-constrained Huffman coding of multiple sources which employs the iterative bisection algorithm. The proposed method provides better performance over conventional allocation methods and can be applied to many adaptive variable-length coders with memory constraints.

Abstract: A data coding system is arranged such that character font data (9) is sampled in 4-bit data units (9a) and these 4-bit data units (9a) are compressed using the Huffman coding system. The data coding system employs a code structure in the Huffman coding system for respectively converting three 4-bit data units selected in order from that of the highest appearance frequency in the character font data (9) into compressed codes expressed using no more than 3 bits. In particular, the data coding system is for use in compressing Chinese character font data (9). The described system permits a significantly improved compression rate of the character font data (9) into respective compressed codes.

Abstract: of EP0745946A data coding system is arranged such that character font data (9) is sampled in 4-bit data units (9a) and these 4-bit data units (9a) are compressed using the Huffman coding system. The data coding system employs a code structure in the Huffman coding system for respectively converting three 4-bit data units selected in order from that of the highest appearance frequency in the character font data (9) into compressed codes expressed using no more than 3 bits. In particular, the data coding system is for use in compressing Chinese character font data (9). The described system permits a significantly improved compression rate of the character font data (9) into respective compressed codes.

Abstract: A quadtree segmented two-dimensional predictive coding technique is presented in this paper for exploiting correlation between adjacent image blocks and uniformity in variable block size image blocks. Visual pattern block truncation coding (VP-BTC) is employed to code the high activity 4/spl times/4 image blocks, with a set of sixteen visual patterns used to encode the 4/spl times/4 bitmaps. Simulation results showed that the new coding technique achieved substantial bit rate reduction with good subjective quality.

Abstract: For memory constrained high-order entropy coding, (i) the number of Huffman tables and (ii) the size of each Huffman table have to be appropriately reduced. Recently, we developed a Huffman table sharing and a memory allocation methods, each of which efficiently reduces either (i) or (ii) to meet the given memory constraint while keeping the increase in average bitrate as small as possible. However, given a memory constraint, the Huffman table sharing and the memory allocation methods have to be employed at the same time to achieve the better result. This paper presents several efficient schemes for combining the two methods along with simulation results.

Abstract: An efficient coding scheme for handwriting and bilevel graphics based on quadtree segmentation and block-run run-length coding (RLC) is presented. By segmenting a captured handwriting or graphic image into variable size blocks according to the image activity inside the blocks and run-length coding on a block-run basis, the proposed scheme is able to achieve higher coding efficiency than the conventional one-dimensional RLC and its bit rate is close to that of on-line dynamographical chain coding.