Abstract: A variable bit rate (VBR) representation of an image sequence is segmented based on a plurality of time intervals. For each of at least two of the time intervals, the method comprises: determining which of the packets, denoted by Pp, in the VBR representation within the presently-considered time interval has a number of blocks of information per unit time greater than a baseline value; determining which of the packets, denoted by Pn, in the VBR representation within the presently-considered time interval has a number of blocks of information per unit time less than the baseline value; and creating a second representation of the image sequence in which some blocks of information Bp are removed from at least one Pp packet and interlaced with blocks of information in at least one Pn packet to produce reformatted packets.

Abstract: Video encoding methods and video encoders that provide improved performance while reducing power consumption. In one aspect, a video encoding method comprises the steps of outputting a parameter for a slice of a current frame, wherein the slice comprises a plurality of macroblocks, processing the slice by consecutively encoding and decoding each macroblock of the slice in response to the parameter, and outputting an interrupt signal for the slice The parameter preferably comprises an address of the first macroblock of the slice, an address of a search area in a previous frame, the search area corresponding to a current macroblock in a current frame, and a number of macroblocks comprising the slice.

Abstract: An adaptive entropy coder is coupled with a localized conditioning context to provide efficient compression of images with localized high frequency variations. In one implementation, an arithmetic coder can be used as the adaptive entropy coder. The localized conditioning context includes a basic context region with multiple context pixels that are adjacent the current pixel, each of the context pixels having an image tone. A state is determined for the basic context region based upon a pattern of unique image tones among the context pixels therein. An extended context region that includes the basic context region is used to identify a non-local trend within the context pixels and a corresponding state. A current pixel may be arithmetically encoded according to a previously encoded pixel having the same tone or as a not-in-context element. In one implementation, a not-in-context element may be represented by a tone in a color cache that is arranged as an ordered list of most recent not-in-context values.

Abstract: A method comprises determining a plurality of time intervals Tp and Tn within a variable bit rate (VBR) representation of an image sequence. The time intervals Tp are those in which a number of blocks of information per unit time is greater than a baseline value. The time intervals Tn are those in which a number of blocks of information per unit time is less than the baseline value. A second representation of the image sequence is created in which some blocks of information Bp are removed from the time intervals Tp and interlaced with blocks of information Bn in the time intervals Tn to reduce a variation in a number of blocks of information per unit time between the time intervals Tp and Tn.

Abstract: A method comprises determining a plurality of time intervals Tp and Tn within a variable bit rate (VBR) representation of an image sequence. The time intervals Tp are those in which a number of blocks of information per unit time is greater than a baseline value. The time intervals Tn are those in which a number of blocks of information per unit time is less than the baseline value. A second representation of the image sequence is created in which some blocks of information Bp are removed from the time intervals Tp and interlaced with blocks of information Bn in the time intervals Tn to reduce a variation in a number of blocks of information per unit time between the time intervals Tp and Tn.

Abstract: A system to decode compressed digital images coded with frequently occurring variable length or other symbols. The disclosed system extracts one or more symbols from the coded compressed image and selects corresponding quantized pixel values for inverse quantization from a table relating the extracted symbols to precalulated quantized pixel values.

Abstract: A method is disclosed for decoding multiple-coded symbols from a coded input symbol stream in a single clock cycle. The method constructs an original Huffman look-up table by extending the associated Huffman tree to decode multiple symbols in one clock cycle in a first embodiment and decodes multiple DCT coefficient symbols in an alternate embodiment. An advantage of the method is that the depth of the new Huffman tree is adjustable thereby making the method easily adaptable to various hardware architectures. A further advantage of the present invention is that the decoding process speed is significantly increased while the size of the lookup table is nominally increased.

Abstract: Conversion between different codes can be implemented at high speed by a small device. A code input unit inputs a code. A compatible information reference unit refers to compatible information of each compressed symbol before and after code conversion, which then sends out compatible information data to a compressed symbol conversion unit. The compressed symbol conversion unit converts each of the compressed symbols in the input code data to a form adapted to a code to be converted based on the compatible information data, which then sends out converted code data to a converted code output unit.

Abstract: Methods and apparatuses for blending two images using vector table look up operations. In one aspect of the invention, a method to blend two images includes: loading a vector of keys into a vector register; converting the vector of keys into a first vector of blending factors for the first image and a second vector of blending factors for the second image using a plurality of look up tables; and computing an image attribute for the blended image using the blending factors.

Abstract: The present invention relates to an apparatus and a method for vector descriptor representation and multimedia data retrieval, which can quantize a plurality of feature values described by a vector descriptor respectively, represent the quantized feature values in the form of bit or orthogonally transform the quantized bector feature values, and rearrange the feature values represented in the form of bit from the highest bit to the lowest bit or rearrange the transformed coefficient from low frequency to high frequency to represent the vector descriptor hierarchically.

Abstract: Image data is decomposed into subbands and quantized by a discrete wavelet transformer (102) and coefficient quantizer (103). On the other hand, a high-resolution region information input unit (105) inputs region designation information H(x, y) that designates a region which requires a high resolution, and a mask generator (106) obtains, based on H(x, y), mask information M(S, x, y) indicating if coefficients of HL2, LH2, and HH2 are generated with reference to pixels of the high-resolution region. A coefficient correction unit (104) corrects the quantized coefficients of respective subbands with reference to M(S, x, y) if subband S is one of HL2, LH2, and HH2, and does not correct if subband S is other than HL2, LH2, and HH2.

Abstract: An image coding apparatus includes an image feature extraction means for generating character data which includes character codes corresponding to character images included in a document image and auxiliary information indicating the sizes and positions of the respective character images in the document image based on image data indicating the document image. The image coding apparatus generates a predictive document image for the document image based on the character data, and subjects the document image data to arithmetic coding in which the probability model is changed with reference to predictive document image data. In the image coding apparatus, the probabilities of white pixels and black pixels indicated by the probability model are very close to those in the character image, whereby the coding efficiency for the character image data in the arithmetic encoder is improved.

Abstract: A method and a system for the compensation of Fixed Pattern Noise (FPN) in digital images have been achieved. The FPN compensation is based on processing done during the production of said images. The fixed pattern noise is here defined as the fixed pattern seen in the individual pixel offsets. The fixed pattern noise is uncorrelated noise but it has a statistical distribution that can be scaled to fit all images. The general idea is to measure the distribution for each individual camera, compress it, and save it in the module. For each image that is then taken with the module the noise pattern can be retrieved and rescaled to fit the image. Covered pixels are employed to normalize the FPN data to the current frame. In order to minimize memory requirements a compression scheme has to be used. A method combining a quantization step with a non-lossy compression is used. The black level is corrected for as part of the operation.

Abstract: Vector quantization provides font contour data reduction. Characters are defined in a font collection using segmented outline forms and common references stored to curve segments for similar segments in different characters of the font collection. Compressing the font collection preferably includes applying an entropy encoding data reduction. The common references are stored in a segment storage location. Another aspect includes decompressing a font store containing a compressed collection of outline fonts. The uncompressed curve segments are scaled so as to match a requested size for a letterform.

Abstract: A method of storing a digital image file, includes the steps of: storing a header defined by a standard; storing image data encoded according to the standard after the header; and storing an image tail containing non-standard meta-data at the end of the image data.

Abstract: A decoding apparatus has: M tables for storing, in correspondence with M types of variable-length code tables, minimum code words or maximum code words of classes of variable-length code words constructing a variable-length code table; a table selector which selects one table from the M tables; N comparators which compare input coded data with the minimum code words or maximum code words outputted from the table selected by the table selector; a switch circuit and a priority encoder which obtain a class number corresponding to an initial code word of the input coded data based on results of comparison by the N comparators; a code length converter which converts the class number into a code length; and an address generator which generates an address to access a memory holding decoded data from the class number and the code length outputted said code length converter. The data outputted from the memory based on the address becomes decoded data of the input coded data.

Abstract: A method and system of lossless compression of integer data using a novel backward-adaptive technique. The adaptive Run-Length and Golomb/Rice (RLGR) encoder and decoder (codec) and method switches between a Golomb/Rice (G/R) encoder mode only and using the G/R encoder combined with a Run-Length encoder. The backward-adaptive technique includes novel adaptation rules that adjust the encoder parameters after each encoded symbol. An encoder mode parameter and a G/R parameter are adapted. The encoding mode parameter controls whether the adaptive RLGR encoder and method uses Run-Length encoding and, if so, it is used. The G/R parameter is used in both modes to encode every input value (in the G/R only mode) or to encode the number or value after an incomplete run of zeros (in the RLGR mode). The adaptive RLGR codec and method also includes a decoder that can be precisely implemented based on the inverse of the encoder rules.

Abstract: Alphabet-partitioned codewords, for example class based codewords, are decoded using a two-part decoder. The first part is a class code decoder that extracts, from the codeword, information for decoding the symbol code. The second part is a symbol code decoder that uses that information to decode the symbol code. If the symbol code is a literal, the symbol is just stripped from the codeword; otherwise, the symbol is retrieved from a symbol memory. Throughput of blocks of even numbers of codewords is enhanced by compressing first and second block halves oppositely and by using two decoders to decompress first and second block halves in parallel. Throughput is enhanced further by using extended codewords, each of which is a concatenation of class codes and symbol codes of several symbols.

Abstract: A method of quantifying perceptual information and entropy for a data input signal uses an appropriate perceptual model to produce a perceptual response. The perceptual response is then entropy converted to produce the perceptual information and entropy. The entropy conversion may include the application of a transform function, such as a Karhunen-Loeve transform function, to the perceptual response to produce a set of coefficients, which coefficients then have probabilities calculated for them based on coefficient histograms of massive tests. The probabilities are used to calculate the perceptual information and entropy using standard techniques. The perceptual information and entropy may be used in applications for achieving greater accuracy subjective quality determinations for the data input signal.

Abstract: A method comprises determining a plurality of time intervals Tp and Tn within a variable bit rate (VBR) representation of an image sequence. The time intervals Tp are those in which a number of blocks of information per unit time is greater than a baseline value. The time intervals Tn are those in which a number of blocks of information per unit time is less than the baseline value. A second representation of the image sequence is created in which some blocks of information Bp are removed from the time intervals Tp and interlaced with blocks of information Bn in the time intervals Tn to reduce a variation in a number of blocks of information per unit time between the time intervals Tp and Tn.

Abstract: Different video codecs can be integrated or duplicated functions of each codec can be shared. According to an apparatus and method for supporting plural codecs of the present invention, a decoding process can be performed by analyzing information of a transmitted bit stream, deciding kind of codec being used by means of a header analysis unit, connecting through a switching unit to a corresponding decoder of a decoding unit including plural decoders or to a corresponding operation block among a plurality of operation blocks including at least one function.

Abstract: An input digital signal of a first format (DV video signal) is restored to a variable-length code by having its framing cancelled by a de-framing section 11, then decoded by a variable-length decoding (VLD) section 12, inversely quantized by an inverse quantizing (IQ) section 13, and inversely weighted by an inverse weighting (IW) section 14. Then required resolution conversion in the orthogonal transform domain (frequency domain) is carried out on the inversely weighted video signal by a resolution converting section 16. After that, the video signal having the resolution converted is weighted by a weighting (W) section 18, then quantized by a quantizing (Q) section 19, coded by a variable-length coding by a variable-length coding (VLC) section 20, and outputted as a digital signal of a second format (MPEG video signal).

Abstract: An apparatus for processing a block of image data at a high speed, in which two items of data consecutive in data scan are simultaneously written in different memories and are processed as a set of data having a valid portion and an invalid portion and in which the subsequent process is differentiated depending on the frequency of occurrence of the data to suppress any increase in the scale of the circuit and to increase the speed of the same.

Abstract: An apparatus for partitioning moving picture data comprises a first quantizing unit for first-quantizing a received video signal and outputting a first-quantized signal; and a second quantizing unit for second-quantizing the first-quantized signal and partitioning the first-quantized signal into a preceding part and a succeeding part.

Abstract: Method and apparatus for use in Huffman decoding are described. In exemplary systems, a JPEG Huffman decoder is capable of simultaneously decoding multiple coefficients and/or symbols in a single table lookup. Methods for designing, building, and using such a table are included. Other embodiments are described and claimed.

Abstract: A variable length encoding unit includes a run-length converter, a table memory, and a variable length encoder. The run-length converter converts block data consisting of a plurality of image signals into a zero-run number and a level value in accordance with a scanning sequence. The table memory stores a VLC (variable length code) and VLC length at an address corresponding to the zero-run number and level value. The variable length encoder reads the VLC and the VLC length from the table memory in response to the zero-run number and level value converted by the run-length converter, and carries out the variable length coding by cutting the VLC from the read data in accordance with the VLC length. The variable length coding unit can flexibly handle various types of variable length coding/decoding schemes including international standard coding methods without insisting on its unique variable length coding.

Abstract: A database managing apparatus, which can immediately obtain desired data from a database. A database managing apparatus includes a CPU, an input/output device, a main memory and an external memory. The CPU has a controller that initially obtains one article record from stored records via the input/output device. The controller reads out definition data in a database definition file. Then, the controller classifies the obtained records according to attributions based on the definition data. The controller does not compress the data regarding the record group belonging to the attribution A, which is the record group to be searched, but compresses data regarding the record groups belonging to the attributions B-E, which are the record groups other than the record group to be searched. As a result, the controller can reduce the unnecessary decompression of record data belonging to other attributions than the attribution to be searched. As a result, the controller can quickly retrieve the requested record.

Abstract: A process for compressing a two dimensional array of data including at least one array block having a plurality of block rows and a plurality of block columns of pixels, each pixel having at least one associated intensity value. The process includes generating a delta frame associated with a pair of the block columns, the delta frame having a plurality of delta rows and delta columns of delta bits, one of the delta rows being a highest order row; determining a number of duplicate rows of the delta rows, each of the duplicate rows forming an identical row byte value having delta bits which have a predetermined relationship with the delta bits of the highest order row; recording information indicative of the number of duplicate rows; and developing compressed data including the information indicative of the number of duplicate rows of the delta frame.

Abstract: The present invention selectively applies one of VLC tables stored in a memory for encoding a coded block pattern of a macroblock according to the number of blocks having an object within the macroblock, the number of blocks obtained using shape information, thereby reducing the amount of data transmitted and increasing coding efficiency. The present invention also selectively applies one of VLD tables stored in a memory for decoding a coded block pattern of a macroblock according to the number of blocks having an object within the macroblock, the number of blocks obtained using shape information.

Abstract: This invention increases the available instruction level parallelism (IPC) of CABAC encoding by decoupling the re-normalization loop and the bit-insertion task required to create the encoded bit-stream. This makes all software implementations of CABAC based encoding significantly faster on digital signal processors that can exploit instruction level parallelism such as very long instruction word (VLIW) digital signal processors. In a joint hardware/software implementation, this invention employs existing Huffman variable length encoding hardware with minimum modifications. The de-coupling of these two tasks of this invention exposes previously hidden underlying instruction level parallelism and task level parallelism.

Abstract: This invention reduces the scale of a Huffman table used for decoding. A queuing unit queues a variable-length code word from a received bitstream. A switch circuit discriminates the type of the code word in accordance with the pattern of a predetermined number of bits at the start of the queued variable-length code word, extracts data having a sufficient code word length from a predetermined bit position on the basis of the discrimination result, and outputs the result to a Huffman table. The Huffman table compares the data from the switch circuit with a variable-length code word stored in advance, and when the data and the variable-length code word coincide, outputs first symbol data. The Huffman table also generates a sum value for the first symbol data, and generates two second symbols from the sum result. A selection unit selects and outputs one of the first symbol and two second symbols in accordance with the type of the received code.

Abstract: This invention proposes to transcode the compressed image, that may be in the JPEG format for example, to an intermediate format that allows pseudo-random access. Such a pseudo-random access would that allow efficient image transformation. By using this format, in most cases a pixel is decoded only once in the entire image transformation process. This is certainly true for the most common transformation operations such as rotation by 90, 180 and 270 degrees. This transcoding would enable image transformations in printers whose memory is insufficient to store the entire decompressed image.

Abstract: A method and device for reducing a compressed image to a target size by reducing the quality of the image by a quality scaling factor. Image statistics inherent to the image are used to compute the size reduction as a function of the quality scaling factor. Using the relationship between the quality and the size of the image, an estimated quality scaling factor is obtained based on the target size in an iterative process until the size reduction corresponding to the estimated quality scaling factor is substantially equal to the target reduction.

Abstract: An image processing apparatus in which the amount of data bits for each of plural block units is controlled to be the same for all block units, so as to facilitate location of each block unit in coded image data without necessarily decoding the entire image. Pixel image data which is formable into plural block units is input for each block unit, and the input image data is orthogonally transformed. The orthogonally-transformed image data is quantized, and variable-length coding is performed on the quantized image data so as to generate variable-length code. The amount of variable length code in each block unit is controlled to be no more than a predetermined amount of data bits, with the predetermined amount being the same for each block unit.

Abstract: The present invention comprises a system for communicating image data and other data. The system comprises a JPEG file having a JPEG data, at least one non-JPEG data, and a computer. The computer is programmed to read a predetermined number of sequential bytes of the non-JPEG data. The computer is also programmed to determine position of each byte in the sequential bytes that contain a JPEG marker and determine a locator byte, which is capable of indicating the position of any bytes in the sequential bytes that contain a JPEG marker. The computer is further programmed to write the locator byte in the JPEG file and write encoded bytes in the JPEG file. The computer writes the encoded bytes in the JPEG file in a manner wherein, for each byte in the sequential bytes, if the byte does not contain a JPEG marker, each byte is replaced with a bit, preferably written as 0, in the order of the byte's occurrence in the sequential bytes.

Abstract: Hierarchical encoding techniques allowing a good compression efficiency and allowing to quickly recognize the outline of an image from a fraction of encoded data. In order to realize such techniques, an image processing apparatus has a generating unit for generating a plurality of coefficients representative of an image, a variable length encoding unit for encoding each of the plurality of coefficients generated by the generating unit, and a hierarchical output unit for distributing each bit of variable length encoded data corresponding to each coefficient and obtained through variable length encoding by the variable length encoding unit, to a plurality of bit planes each corresponding to a level of each bit, and hierarchically and sequentially outputting the plurality of bit planes.

Abstract: An image processing unit is provided which can solve a problem of a conventional image processing unit in that the control for coding image data is complicated and its circuit scale is large. The image pickup apparatus picks up an image of a subject and converts it to electric signals, thereby outputting image signals. An A/D converter A/D converts the image signals, and outputs them as image data. A fixed length coding circuit breaks down the image data into small unit blocks consisting of a predetermined number of pixels, and carries out fixed length coding after obtaining the average level of the pixel data in each unit block. An exposure controller calculates a luminance level of a whole set of pixel data in the unit blocks, and controls the exposure of the image pickup apparatus so that image data becomes a predetermined luminance level.

Abstract: This invention relates to compression of prediction error in motion compensated video coding schemes. Prediction frame and motion vectors are used to extract useful information about the properties of the prediction error signal, e.g. its location, directionality, etc. This information does not need to be transmitted because the decoder of the proposed system includes means to extract this information. Information extracted by the encoder and decoder is used to adapt prediction error coding scheme which gives substantial bitrate reductions.

Abstract: Decoding variable length codes having regular bit pattern prefixes enables faster decoding of variable length codes, especially in systems that provide bit or bit mask search capabilities. An embodiment of the present invention determines a code prefix type, and calculates a length of the code prefix. A first data structure may be provided to associate the maximal number of bits in a variable length code with the length of the code prefix, and to locate further decoding data in accordance with the prefix length and type. A bit stream may be read according to the maximal length obtained. An additional data structure may be provided to retrieve a decoded value and the actual length of a variable length code being decoded. This data structure may be indexed with the value of the bit combination read from the bit stream. In case the actual length of the variable length code is less than the maximal length, the excess bits may be returned to the bit stream.

Abstract: When a plurality of image data are to be compressed, an image correlation information extraction unit extracts respective image correlation information of image data, using first image data as reference image data. An encoding unit compresses the extracted image correlation information to produce encoded data. A data output unit outputs the compressed data and the reference image data to a page memory. When the compressed data are to be restored, a data input unit supplies the image correlation data input from the page memory to an image correlation information decoding unit, and supplies the reference image data to an image restoring unit. The image correlation information decoding unit decodes the image correlation data into the respective image correlation information and supplies the image correlation information to the image restoring unit. The image restoring unit restores the plural image data from the respective image correlation information and the reference image data.

Abstract: The object-oriented coder discriminates resource allocation between objects and non-objects for video messaging applications over wireless networks. The object-oriented coder executes a rate control algorithm, an unequal error protection algorithm, and an error concealment algorithm. In the rate control algorithm, an iterative feedback rate control scheme is used in which quantization values of object and non-object data are held constant. In the unequal error protection algorithm, the bit stream is partitioned by object macroblocks and non-object macroblocks. In the error concealment algorithm, five bits of QUANT values of each GOB are used for representing location and motion vectors of the object in the next frame, since the quantization value is constant. The five bits are not used for quantization value. The five bits are used for error concealment to avoid bit rate overhead. The object-oriented coder increases encoding delay, but this increase is acceptable in messaging.

Abstract: A fast loss less image compression system based on neighborhood comparisons compares pixel value differences with neighboring pixels and replaces such pixel values with the minimum of the differences. A marker is attached to a block of pixels, such that all the pixels in that block are compared with neighbors of one direction. The marker indicates how all of the pixels in that block are compared. Intermittent Huffman-tree construction is used such that one tree is used for several frames. Huffman coding is used to compress the resulting frame. A single Huffman-tree is constructed once every predetermined number of frames. The frequency of Huffman-tree construction can be performed according to the instantaneous availability of processor time to perform the construction. When more processing time is available, the Huffman-trees are computed more frequently. Such frequency variation can be implemented by using an input video frame buffer.

Abstract: A method of performing variable-length coding for a multilevel image with simple processing, and a processing apparatus for executing this variable-length coding method at a high speed have not been established yet. According to this invention, discrete wavelet transformation processing is performed for multilevel image data to quantize the image data in units of subblocks of respective frequency components. An optimal shift parameter is easily determined at a high speed for each subblock. This realizes high-speed variable-length coding based on the parameter.

Abstract: A coded data length detection apparatus and method capable of detecting a coded data length at a high accuracy at a high speed, wherein original image data input to an image compression and coding circuit is processed by DCT at a DCT unit and quantized at a quantizer, quantization coefficients at this time is obtained by switching first set of scale coefficients and second set of scale coefficients in a lattice pattern in units of MCU of an image in scale coefficients switch, Huffman coding is performed on a quantized result at a Huffman coding unit, a data length is counted at a first compressed data length counter or a second compressed data length counter, and the counted data lengths is respectively multiplied by 2× at a first double multiplier or a second double multiplier to find a data length at the time of encoding the entire image by the respective scale factors.

Abstract: An adaptive entropy coder is coupled with a localized conditioning context to provide efficient compression of images with localized high frequency variations. In one implementation, an arithmetic coder can be used as the adaptive entropy coder. The localized conditioning context includes a basic context region with multiple context pixels that are adjacent the current pixel, each of the context pixels having an image tone. A state is determined for the basic context region based upon a pattern of unique image tones among the context pixels therein. An extended context region that includes the basic context region is used to identify a non-local trend within the context pixels and a corresponding state. A current pixel may be arithmetically encoded according to a previously encoded pixel having the same tone or as a not-in-context element. In one implementation, a not-in-context element may be represented by a tone in a color cache that is arranged as an ordered list of most recent not-in-context values.

Abstract: A coded image data decoding system is designed to depack and variable-length decode image data packed by an image compressing/decompressing format wherein the sum of the length of components other than variable-length code components in one block and the length of an end code is not shorter than a bit length obtained by subtracting 1 bit from the number of bits of the maximum length of variable-length code words.

Abstract: This invention has as its object to attain efficient encoding/decoding even when the processing time, memory, arithmetic cost, and the like of an apparatus are limited. Encoded image data is input to an image processing apparatus (1) to obtain transform coefficients of subbands. The transform coefficients required to decode an image from its head position line by line are encoded by Golomb coding using one line of a subband as a unit to generate a code sequence, which is output to an image output apparatus (2). The image output apparatus (2) reconstructs an image by decoding the Golomb encoded data, dequantizing the coefficients, and computing the inverse discrete wavelet transforms of the coefficients, and outputs the reconstructed image.

Abstract: A surveillance image is acquired from a surveillance camera. Supplement information (such as date and time, and surveillance camera number) relating to the surveillance image is acquired. By imaging the supplement information, supplement information image data is generated. A quantized discrete cosine transform coefficient block obtained by conducting discrete cosine transform and then quantization on surveillance image data of the surveillance image is added to a quantized discrete cosine transform coefficient block obtained by conducting discrete cosine transform and then quantization on the supplement information image data. A resultant sum is subjected to Huffman encoding. The result is recorded.

Abstract: Encoded data using reversible variable length code words is input to a forward decoder (123) to be decoded in the forward direction. When an error is detected in the encoded data in the forward decode processing, backward decode processing is started by a backward decoder (126). A decode value determination unit (125) determines a decode value by using the forward and backward decode results and the error detection positions in the encoded data in units of bits and syntax which are respectively detected in the forward decoding and the backward decoding.

Abstract: Variable-length encoder 16 feeds a CBP as an input symbol H1 to variable-length code output part 30. Coded symbol memory 31 supplies CBPs in neighboring blocks as coding map table reference information H2 to coding map table provider 32. Coding map table provider 32 determines a coding map table used in coding of the CBP, based on these CBPs in the neighboring blocks, and provides a coding map table H4 to variable-length code output part 30. Variable-length coding table provider 33 feeds a variable-length coding table H5 to variable-length code output part 30. Then the coding target CBP is subjected to variable-length coding, and the resultant is outputted as coded data D9. This allows information source coding of coding symbols to be performed efficiently according to the coding condition and the property of image.