Abstract: This present invention provides an image processing system and image processing method which can reliably transmit image information to a destination without attaching a large file which applies load to an e-mail system or reception terminal and make the receiving side easily acquire necessary image data on the basis of determination on the receiving side. In an image input/output device (10), image information is input from an image input device (201) and stored in a HDD (208) in a control unit (200). A low-resolution image or vector data is generated from the image information in accordance with the properties of objects contained in the image information. The generated information and information about the storage location of the image information are transmitted to a designated transmission destination.

Abstract: Methods, apparatus and computer readable medium are described that compress and/or decompress digital images in a lossless or a lossy manner. In some embodiments, a display controller may quantize pels of a digital image and may identify runs of successive quantized pels which are equal. The display controller may generate a symbol to represent an identified run of pels. The symbol may comprise a run length and a quantized pel that may be used to reconstruct the run of pels. The symbol may further comprise an error vector for each of the pels of the run that may be used to further reconstruct the run of pels.

Abstract: A process for secured distribution of fixed numerical images according to a nominal format resulting from numerical encoding in wavelets, represented by a original stream including a packet relating to organization of a binary sequence that contains at least a block that regroups numerically encoded simple elements according to a mode specified inside the stream and utilized by decoders that are capable of reconstructing or decoding it to be able to correctly display the image including modifying at least one of the simple elements according to at least a substitution operation including extracting the simple element, followed by its replacement by lure data, modifying a principal stream to conform to a nominal format including modified blocks and packets, and by a path that is separate from the principal stream of complementary numerical information and allowing reconstruction of the original stream from calculations, on destination equipment, as a function of the principal stream and the complementary infor

Abstract: A selector (502) for selecting a background motion vector for a pixel in an occlusion region of an image, from a set of motion vectors being computed for the image, comprises: computing means (510) for computing a model-based motion vector for the pixel on basis of a motion model being determined on basis of a part of (402-436) a motion vector field (400) of the image; comparing means (511) for comparing the model-based motion vector with each of the motion vectors of the set of motion vectors; and selecting means (512) for selecting a particular motion vector of the set of motion vectors on basis of the comparing and for assigning the particular motion vector as the background motion vector.

Abstract: Systems and methods for image pattern recognition comprise digital image capture and encoding using vector quantization (“VQ”) of the image. A vocabulary of vectors is built by segmenting images into kernels and creating vectors corresponding to each kernel. Images are encoded by creating a vector index file having indices that point to the vectors stored in the vocabulary. The vector index file can be used to reconstruct an image by looking up vectors stored in the vocabulary. Pattern recognition of candidate regions of images can be accomplished by correlating image vectors to a pre-trained vocabulary of vector sets comprising vectors that correlate with particular image characteristics.

Abstract: To always maintain good image quality no matter whether the target bit rate for encoding is high or low. An encoder 210 includes: a converting section 103 for converting picture data representing a moving picture into a coefficient including a plurality of frequency components on a predetermined unit basis; a determining section 212 for determining a matrix to define a frequency bit allocation characteristic based on the resolution of the moving picture and a target bit rate for encoding that has been specified in advance; a quantization section 204 for quantizing each coefficient based on the matrix determined, thereby generating a quantized value; and an encoding section 105 for encoding the quantized value to generate encoded data of the moving picture.

Abstract: A technique wherein the number and position of a quantization parameter node is determined in response to the quantization parameters and a preselected error. The size of scene graph and the corresponding amount of memory required to store the scene graph can be reduced by selective placement of quantization parameter nodes in a scene graph. The scene graph is traversed depth first to establish an order and then traversed in reverse. At each node, a calculation relating to (1) the relative cost of inserting a quantization parameter node and (2) the relative savings that result from insertion of a quantization node is performed. Quantization parameter nodes are selectively placed in response to a result of these calculations. The maximum degree of acceptable error value is chosen for each quantization type. This error value limits the number of quantization parameter nodes that can be placed in a scene graph.

Abstract: Methods, systems, and computer programs for encoding images are described. In one aspect, quantized frequency domain vectors are sequentially generated from a sequence of blocks of the image. Each quantized frequency domain vector includes a set of quantized forward transform coefficients that are derived from a respective image block. For each successive quantized frequency domain vector, a current input capacity level of a buffer is determined and the quantized frequency domain vector is modified to increase compressibility when the current input capacity level is determined to be below a prescribed threshold. Modified and unmodified quantized frequency domain vectors are encoded into a sequence of encoded image blocks. The sequence of encoded image blocks is stored in the buffer.

Abstract: A data compression system and method provides a best base searching method to determine the best orthogonal basis function for wavelet-based, data compression. A processing device may receive data signals from a source and wavelet-based data compression may be performed on the received data before transmission (to a receiving component) and/or further digital signal processing (DSP) of the data signals. An encoder portion of the processing device performs the data compression using a predetermined algorithm that determines the best orthogonal basis function (base) of the signal transform (used for representing the data signals) by searching a set of bases including (approximately) all orthogonal bases available to provide the minimum number of coefficients for efficient data compression while maintaining a low error rate in reconstructing the original data signals.

Abstract: A vector error diffusion (VED) method employable in cycles with respect to a hi-tonal color printing engine which prints bi-tonal color images in a device output color space. The method generally includes (a) acquiring input color-image data which is characterized with an input color space, (b) processing, with available pre-established VED accumulated error data, such input data to produce a VED-processed input color-image data stream, (c) from such VED-processed input color-image data stream, creating, without employing interpolation, a VED-processed output color-image data stream which is characterized by the mentioned device output color space, and which is suitable for delivery to and use by the mentioned printing engine, and (d) changing, as appropriate for the next cycle, the VED accumulated error data which will be employed in that next cycle as pre-established VED accumulated error data.

Abstract: A method and system separates components in individual signals, such as time series data streams. A single sensor acquires concurrently multiple individual signals. Each individual signal is generated by a different source. An input non-negative matrix representing the individual signals is constructed. The columns of the input non-negative matrix represent features of the individual signals at different instances in time. The input non-negative matrix is factored into a set of non-negative bases matrices and a non-negative weight matrix. The set of bases matrices and the weight matrix represent the individual signals at the different instances of time.

Abstract: A method, system, and computer program product are provided, wherein the bandwidth necessary to transmit an item of image data is reduced. When items of image data are identified in a datastream, they are extracted from the datastream. The image data item is then divided into a series of subregions of variable size. Where efficiency dictates that the operation is appropriate, the subregions are replaced in the image data item with a unique identifier to produce a reduced image. The reduced image is then packaged into a new data structure containing a header, the reduced image, and a decoding table that will allow the replacement of the identifiers with the extracted subregions. Where subregions are repeated, as they frequently are in images of large size, this arrangement will allow for the compression of the image by the elimination of redundant data that merely represents a repeated subregions. When the image reaches its destination, the it is decoded to reproduce the original image.

Abstract: Stored digital data is searched for on the basis of an input image, difference information is extracted by comparing the retrieved digital data and the input image, and the difference information is composited to the digital data. Digital data generated by composition is stored. When no digital data is retrieved, the input image is converted into vector data, and the image that has been converted into the vector data is stored as digital data. Obtained region segmentation information and an input image are composited, the composite image is displayed on an operation screen of an MFP, and a rectangular block to be vectorized is designated as a specific region from the displayed region segmentation information. A user designates the specific region by designating rectangular blocks in an image using a pointing device.

Abstract: According to the invention, quantization encoding is conducted using the probability density function of the source, enabling fixed, variable and adaptive rate encoding. To achieve adaptive encoding, an update is conducted with a new observation of the data source, preferably with each new observation of the data source. The current probability density function of the source is then estimated to produce codepoints to vector quantize the observation of the data source.

Abstract: This device includes an image data creator that creates a first type of image data for test shooting and creating a second type of image data for real shooting; a compressor that compresses the image data in a predetermined compression format; and a processor; wherein the compressor has one or plural compression parameters relating to a compression rate; the creator newly supplies files of the first type to the compressor one after another; the compressor compresses at least two files among the continuously supplied first type using values of compression parameters; the processor decides the value of the compression parameter to be used for the second type according to a predetermined standard, based on two and more files of first type image data that are compressed using values of compression parameters; and the processor sets the value of the compression parameter of the compressor at the decided value.

Abstract: An image distribution system has a source that encodes digital images and transmits them over an error-prone channel to a destination. The source has an image coder that processes the digital images using vector transformation followed by vector quantization. This produces groups of vectors and quantized values that are representative of the images. The image coder orders the vectors in the codebooks and assigns vector indexes to the vectors such that a bit error occurring at a less significant bit in a vector index results in less distortion than a bit error occurring at a more significant bit. Depending upon the format and the capabilities of the source and destination, the image coder may allocate different numbers of bits to different groups of vectors according to a bit allocation map for this allocation process. The source also has a UEP (Unequal Error Protection) coder that layers the vector indexes according to their significance.

Abstract: A system and method are disclosed which include providing a first layer of an image, the first layer including features of the image having locations within the first layer; and providing a second layer of the image, the second layer including data blocks corresponding to respective ones of the features; each data block being in a location in the second layer substantially corresponding to a location in the first layer of the feature corresponding to each data block, wherein a size and shape of the second layer substantially correspond to a size and shape of said first layer.

Abstract: A method and apparatus for efficiently encoding images using a set of non-orthogonal basis functions, thereby allowing reduction of file size, shorter transmission time, and improved accuracy. The non-orthogonal basis functions include homogenous color basis functions, luminance-encoding basis functions that have luminance edges and chromatic basis functions that exhibit color opponency. Some of the basis functions are non-orthogonal with respect to each other. Using these basis functions, a source vector is calculated to provide a number of coefficients, each coefficient associated with one basis function. The source vector is compressed by selecting a subset of the calculated coefficients, thereby providing an encoded vector. Because the method is highly efficient, the image data is substantially represented by a small number of coefficients. In some embodiments, the non-orthogonal basis functions include two or more classes. A wavelet approach can also be utilized.

Abstract: A technique wherein the number and position of a quantization parameter node is determined in response to the quantization parameters and a preselected error. The size of scene graph and the corresponding amount of memory required to store the scene graph can be reduced by selective placement of quantization parameter nodes in a scene graph. The scene graph is traversed depth first to establish an order and then traversed in reverse. At each node, a calculation relating to (1) the relative cost of inserting a quantization parameter node and (2) the relative savings that result from insertion of a quantization node is performed. Quantization parameter nodes are selectively placed in response to a result of these calculations. The maximum degree of acceptable error value is chosen for each quantization type. This error value limits the number of quantization parameter nodes that can be placed in a scene graph.

Abstract: A blind watermarking method by grouping codewords for VQ-quantized images is disclosed. Especially there is provided a watermark insertion method in which a codebook is divided into three groups satisfying specific standards and a codeword is allocated based on the group to which the corresponding codeword belongs to insert watermark information, and a watermark extraction method for extracting the watermark inserted by the watermark insertion method. According to the present invention, watermark information can be uniformly inserted into an input vector space and the original image is not needed for extracting the watermark information.

Abstract: Systems and methods for image pattern recognition comprise digital image capture and encoding using vector quantization (“VQ”) of the image. A vocabulary of vectors is built by segmenting images into kernels and creating vectors corresponding to each kernel. Images are encoded by creating a vector index file having indices that point to the vectors stored in the vocabulary. The vector index file can be used to reconstruct an image by looking up vectors stored in the vocabulary. Pattern recognition of candidate regions of images can be accomplished by correlating image vectors to a pre-trained vocabulary of vector sets comprising vectors that correlate with particular image characteristics.

Abstract: A system and method are disclosed which include providing a first layer of an image, the first layer including features of the image having locations within the first layer; and providing a second layer of the image, the second layer including data blocks corresponding to respective ones of the features; each data block being in a location in the second layer substantially corresponding to a location in the first layer of the feature corresponding to each data block, wherein a size and shape of the second layer substantially correspond to a size and shape of said first layer.

Abstract: The present invention relates to a real-time wideband compressor for multi-dimensional data. The compressor comprises a plurality of compression engines for simultaneously compressing a plurality of data subsets of a set of input data vectors and providing compressed data thereof using one of SAMVQ or HSOCVQ data compression. Each compression engine comprises an along spectral vectors codevector trainer as well as an across spectral bands codevector trainer. The compression engines are programmable to perform either along spectral vectors codevector training or across spectral bands codevector training in combination with one of the SAMVQ or HSOCVQ techniques without changing hardware. The compressor further comprises a network switch for partitioning the set of input data vectors into the plurality of data subsets, for providing each of the plurality of data subsets to one of the plurality of compression engines, and for transmitting the compressed data.

Abstract: A method of compressing an image is described in which digital data signals in a 2-dimensional images are formed into an image data pyramid with a number of layers and each layer is processed to give a compressed encoding in an ordered list. The encoding with the largest quality gain factor is selected first and added to a compressed representation of the data array. This is repeated for the next largest gain factor and so on until a predetermined maximum is reached. Each layer of the image data pyramid corresponds to different frequency bands, the vector quantizations of these layers will only minimally interfere with one another. This allows a simple ordering of all possible gain contributions made by the compressed encodings, to the compressed representation. This in turn allows a straightforward selection of the compressed encodings having the largest quality gain factors, for compiling the compressed representation of the image.

Abstract: In an encoding method of moving pictures which generates a predictive picture for a current picture based on a reference picture and a motion vector, a macroblock is divided into subblocks. In each of the plurality of subblocks, an initial value of the motion vector is set and an evaluated value E on a difference between the current picture and the reference picture is calculated along a steepest descent direction to determine the minimum value. Then, the smallest evaluated value is selected among the minimum values obtained on the plurality of subblocks to determine the motion vector based on the pixel position of the smallest value.

Abstract: According to the invention, quantization encoding is conducted using the probability density function of the source, enabling fixed, variable and adaptive rate encoding. To achieve adaptive encoding, an update is conducted with a new observation of the data source, preferably with each new observation of the data source, preferably with each new observation of the data source. The current probability density function of the source is then estimated to produce codepoints to vector quantize the observation of the data source.

Abstract: In an image decoding method, a DC image, corresponding to each block mean value per B pixel from encoding data with respect to the HVQ system, is reproduced. A part of the DC image is made a DC nest. Image data of a target block is reproduced by synthesizing, to the DC value of target block, one or more basis vectors selected from DC nests based on the encoding data. The method is improved by setting the lowest n (n?log2 B) bits of the DC pixels in each sample to 0, and down-sampling the selected block from the DC nest and calculating the block mean value using the samples.

Abstract: A procedure for fast training and evaluation of support vector machines (SVMs) with linear input features of high dimensionality is presented. The linear input features are derived from raw input data by means of a set of m linear functions defined on the k-dimensional raw input data. Training uses a one-time precomputation on the linear transform matrix in order to allow training on an equivalent training set with vector size k instead of m, given a great computational benefit in case of m>>k. A similar precomputation is used during evaluation of SVMs, so that the raw input data vector can be used instead of the derived linear feature vector.

Abstract: A procedure for fast training and evaluation image classification systems using support vector machines (SVMs) with linear input features of high dimensionality is presented. The linear input features are derived from raw image data by means of a set of m linear functions defined on the k-dimensional raw input data, and are used for image classification, including facial recognition tasks.

Abstract: Compression and reconstruction of a digital image are both performed by accessing a plurality of color caches corresponding to different chromatic contexts, selecting a color cache for a pixel value being processed, and using information in the selected color cache to predict a value for the pixel being processed.

Abstract: A method of introducing a non-perceptional signal (watermark) to a digital media data is disclosed. The method is based on the representation of source digital data using a special matrix, insertion of a digital watermark into the special matrix to receive the watermarked matrix, and generation of the watermarked data using the source data and the watermarked matrix. In addition, watermark detection of the watermarked data is performed by calculating the special matrix from the watermarked data.

Abstract: Various techniques and tools for chrominance motion vector rounding are described. For example, during motion compensation, a video encoder or decoder converts luminance motion vectors into chrominance motion vector having quarter-pixel accuracy. Or, during motion compensation, a video encoder or decoder converts luminance motion vectors into chrominance motion vectors using one of multiple available chrominance motion vector rounding modes.

Abstract: This invention relates to a method and apparatus for image classification. More particularly, the present invention provides a technique to classify an image as a picture, a graphic or a mixed mode image. The classification is based on an approximation of a segmentation. The approximation is HVQ-LUT-based and outputs classification maps indicating whether pixels are background, text or pictures. Said classification maps are filtered to eliminate odd isolated samples and the resulting count of picture, text and background pixel is analyzed before concluding whether the image has pictorial, graphical or mixed contents.

Abstract: A method of compressing an image frame composed of an array of pixels in the form of digital signals comprises a two stage codebook search. In the first stage the pixelated image frame and the pixelated codebook patches are transformed to frequency domain coefficients and a pattern comparison is made between the coefficients of the image patch and the codebook patches to identify a short list of possible match codebook patches by discarding incorrectly matching patterns. In the second stage the image frame is compared with the short list of codebook patches by any desired method, e.g. by pixel comparisons, to select the best matching codebook patch.

Abstract: A method for processing color-image pixel-data for outputting by a multi-bit (N) color-image output device including the steps of (a) characterizing color-image pixels into N successive, associated scan lines of pixels, which lines are bridged, effectively for processing purposes, by successive N×N pixel groups distributed along the associated scan lines, and (b) for each such pixel group, applying processing to the pixels in the group, including vector error diffusion processing, in a pixel-by-pixel and scan-line-by-scan-line manner employing, cyclically, N different color palettes, one each for the different scan lines—one of which palettes is bi-tonal, and other(s) of which is (are) of a higher bit level up to the level N.

Abstract: A method and apparatus for color quantization of an image employs a dynamic color map. Pixels of a first image are mapped into corresponding color space regions while incrementing counts in records of those color space regions. Associated colors of color space regions having largest non-zero counts are included in a color map. If the number of such associated colors is less than a maximum number of color indexes of the color map, the unassigned color indexes are reserved for later assignment. Pixels of subsequent images are also mapped into corresponding color space regions while incrementing counts in records of those color space regions. If unassigned color indexes are available in the color map, associated colors of these color space regions are added to the color map until all color indexes have been assigned.

Abstract: A data compression system is provided in accordance with the present invention. The system includes a scanning component which scans at least a portion of a transformed image. The scan is performed substantially in a horizontal direction on a first section of the portion and in a vertical direction on a second section of the portion to enable improved data compression of the transformed image. The horizontal and vertical scan directions are performed via a contiguous scan of the respective sections to further enable improved data compression of the transformed image.

Abstract: A face recognition system and method project an input face image and a set of reference face images from an input space to a high dimensional feature space in order to obtain more representative features of the face images. The Kernel Fisherfaces of the input face image and the reference face images are calculated, and are used to project the input face image and the reference face images to a face image space lower in dimension than the input space and the high dimensional feature space. The input face image and the reference face images are represented as points in the face image space, and the distance between the input face point and each of the reference image points are used to determine whether or not the input face image resembles a particular face image of the reference face images.

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: Methods and apparatus implementing techniques for identifying, in a device space, an effective centerscan object color along an edge between an overscan object and a centerscan object, the overscan object having a higher paint order than the centerscan object. The edge is mapped to the device space. A set of overscan boundary pixels is identified in the device space, the overscan boundary pixels being device space pixels that are intersected by the edge. A vector pointing in a direction of the centerscan object relative to the edge is created. The vector is applied to each overscan boundary pixel in the set of overscan boundary pixels to identify a corresponding set of centerscan boundary pixels in the device space. Each centerscan boundary pixel is mapped to the centerscan object to identify a color of the centerscan boundary pixel. A corresponding method for reversed paint order is also described.

Abstract: An image processing system including an image encoder and image decoding system is provided. The image encoder system includes an image decomposer, a block encoder, and an encoded image composer. The image decomposer decomposes the image into blocks. The block encoder, which includes a selection module, a codeword generation module and a construction module, processes the blocks. Specifically, the selection module computes a set of parameters from image data values of a set of image elements in the image block. The codeword generation module generates codewords, which the construction module uses to derive a set of quantized image data values. The construction module then maps each of the image element's original image data values to an index to one of the derived image data values. The image decoding system reverses this process to reorder decompressed image blocks in an output data file.

Abstract: A technique wherein the number and position of a quantization parameter node is determined in response to the quantization parameters and a preselected error. The size of scene graph and the corresponding amount of memory required to store the scene graph can be reduced by selective placement of quantization parameter nodes in a scene graph. The scene graph is traversed depth first to establish an order and then traversed in reverse. At each node, a calculation relating to (1) the relative cost of inserting a quantization parameter node and (2) the relative savings that result from insertion of a quantization node is performed. Quantization parameter nodes are selectively placed in response to a result of these calculations. The maximum degree of acceptable error value is chosen for each quantization type. This error value limits the number of quantization parameter nodes that can be placed in a scene graph.

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: An image processing system including an image encoder and image decoding system is provided. The image encoder system includes an image decomposer, a block encoder, and an encoded image composer. The image decomposer decomposes the image into blocks. The block encoder, which includes a selection module, a codeword generation module and a construction module, processes the blocks. Specifically, the selection module computes a set of parameters from image data values of a set of image elements in the image block. The codeword generation module generates codewords, which the construction module uses to derive a set of quantized image data values. The construction module then maps each of the image element's original image data values to an index to one of the derived image data values. The image decoding system reverses this process to reorder decompressed image blocks in an output data file.

Abstract: Briefly, in accordance with one embodiment on the invention, a method of compressing a data set includes the following. In multiple passes, each data signal in the data set is categorized into a category of a predetermined set, and, for selected categories of the predetermined set, the data signals for that category are coded using a codebook for that category. Briefly, in accordance with another embodiment of the invention, a method of decompressing a compressed data set includes the following. For compressed data signals in the data set in one category of a predetermined set of categories, a signal associated with the particular category is employed for the compressed data signal, and, for selected categories of the predetermined set, the compressed data signals for that category are decoded using a codebook for that category.

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: This invention relates to the creation of dictionary functions for the encoding of video signals using matching pursuit compression techniques. After an initial set of reference dictionary images is chosen, training video sequences are selected, and motion residuals are calculated. High energy portions of the residual images are extracted and stored when they match selection criteria with the reference dictionary. An energy threshold is used to limit the number of video signal “atoms” encoded for each frame, thus avoiding the encoding of noise. A new dictionary is then synthesized from the stored portions of the image residuals and the original reference dictionary. The process can then be repeated using the synthesized dictionary as the new reference dictionary. This achieves low bit rate signals with a higher signal-to-noise ratio than have been previously achieved.

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: 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: An apparatus for removing a shade from an image includes a principle component analyzing unit performing principle component analysis on a set of face images having various types of shades to generate an eigen space, a storing unit storing the eigen space generated by the principle component analyzing unit, an image input unit receiving as an input a new face image, a projecting unit projecting the face image input through the image input unit to the eigen space stored in the storing unit, and an image producing unit producing a face image with the shade component removed, based on the face image input through the image input unit and the image projected to the eigen space by the projecting unit.