Abstract: A method and apparatus for parallel processing of data without the need for cross-communication or synchronization between processing nodes is provided. The system is especially suitable for use in a wafer inspection system for the semiconductor industry. Embodiments include scanning a small area of the wafer, and distributing pixel data among a plurality of processing nodes, each of which accept and process an optimal amount of pixel data independently of the other processing nodes, thereby enabling increased throughput without increasing system complexity.

Abstract: A method and an apparatus for fault-tolerant computing in a parallel processing environment. Intermediate data points are determined from a set of input data by a plurality of processing elements operating in parallel. Preferably, the resolution of the intermediate data points is greater than the resolution required in an output imaging data set. The intermediate data points are spatially re-gridded to the resolution required in the output imaging data set, ignoring any missing data as a result of a processing element failing.

Abstract: Method and device for extracting a specified image subject including an extracting condition of an extracting algorithm of a subsequent stage so as to be adapted to an extraction result by the extracting algorithm of a precedent stage, when specified image subject extracting algorithms are successively implemented. At the time of performing the specified image subject extracting algorithms in each stage by parallel processing, the method and device manage extraction states of each step and qualify extraction processing conditions in the subsequent stage in accordance with the extraction states in the precedent stages. The method and device perform a vote in an N-dimensional space of image characteristic quantity for each extraction area by the specified image subject extracting algorithm and then perform weighting of degree of certainty as the specified image subject based on an aggregation value of the vote within a section area for aggregation in the N-dimensional space.

Abstract: A printer performs printing after appropriately correcting irregularities, thin-line patchiness, and isolated pixels. A first image-quality corrector unit detects slanting-line irregularities represented by the black and white pixel data generated through binary processing according to a method other than the error-variance method to thereby perform smoothing processing therefor. A second image-quality corrector unit detects defects such as irregularities and patchiness specific to the binary processing according to the error-variance method to thereby perform smoothing processing therefor. A third image-quality corrector unit detects isolated pixels in gray fields that are specific to the binary processing according to the error-variance method to thereby distribute the isolated pixels to peripheral pixels.

Abstract: In a method for controlling mechanisms or technical systems, the mechanisms or technical systems to be controlled are stored in a controller with their states, and with associated signal formers of sensors and actuators, whereby starting from a defined reference state at the onset of the activation of the controller, the actual states signaled by the technical system via the sensors are continuously compared with the specified state, the specified state being stored in the controller, and, based on this comparison, every deviation from the specified state is identified in the technical system, and, when initiated, a new instruction that changes the state of the mechanisms or of the technical system updates the specified state for the comparison and monitors the time till the acknowledgment of the new state, and sensor signals and comparable information exclusively serve the state identification of elementary functions and state changes exclusively ensue upon the initiation of elementary instructions.

Abstract: First and second reading modules read first and second signals in the main scanning direction of the image of a document. The scanning module includes a carriage for relatively moving the first and second reading modules and the image of the document in a sub-scanning direction to make the modules scan the image. A scanning-control-condition selecting module selects scanning control conditions of the scanning module in accordance with a plurality of read magnifications preset as read magnifications of the document by the first and second reading modules. A first setting module sets the scanning control conditions to a plurality of control conditions in accordance with the read magnification of the first reading module. A second setting module sets the scanning control conditions to a plurality of control conditions in accordance with the read magnification of the second reading module.

Abstract: A digital microelectronic circuit comprises a clocked data-processing unit (1) and a converting unit (2) which reads in data present at the output of the data-processing unit, performs a predetermined converting operation on the data and passes on the converted data. The converting unit is realized in an asynchronous logic circuit, such that the period of time for performing the converting operation is shorter than the shortest time interval to the next change of the data present at the output of the data-processing unit. In this way, fast, serial synchronous processes can be parallelized from the point of view of the slow synchronous system in synchronous systems which are slow relative thereto by using asynchronous logics, without a further high-frequency clock system being required.

Abstract: The image-processing apparatus includes n image processing sections which receive n consecutive pixel data items that are respectively input with the same timing and which respectively process the respective input pixel data items with the same timing, and a control section that controls the n image processing sections. Each of the image processing sections are capable of being set to one of a first operation mode allowing data communication with the controlling section and a second operation mode allowing only reception from the aforementioned controlling section, one of the image processing sections is set to the first operation mode, and n?1 of the image processing sections are set to the second operation mode.

Abstract: A method is provided of processing data representing pixel color having a luminance component and color difference components. The data is divided into first and second data portions, the first data portion comprising the luminance components and the second data portion comprising the color difference components. First and second instructions from a combined instruction word, and the first and second date portions are processed in parallel using first and second parallel processors within a processor architecture, the first and second parallel processors operating according to the first and second instructions, respectively. The processed first and second data portions are combined to provide processed pixel color data. This method uses parallel processor sections to process the luminance and color difference components. The parallel processor sections can then use instructions suited to the type of data being processed, providing an efficient method of processing the graphics data.

Abstract: Techniques for removing ringing artifacts from video data. A deringing filter in accordance with the present invention preserves real image edges in a video frame, while smoothing out the interiors of objects. In one aspect, a 9-tap low-pass filter is applied to an adaptive processing window. The filter window is initialized with the values in a 3×3 mask centered on the position whose output is computed. Then all values that are very different from the central one are replaced with the central value. The deringing filter varies between 3×3 low-pass and identity, depending on how much the central value differs from its surrounding ones. A deblocking filter in accordance may also be suitably used in conjunction with the deringing filter.

Abstract: An appearance inspection apparatus is composed of a memory 14, a thread generator and a plurality of CPUs 10 to 13. The memory 14 stores image data of an appearance of an IC. The thread generator generates a thread in which a procedure is described for independently processing the image data stored in the memory 14 and storing the processing result into the memory 14. The plurality of CPUs 10 to 13 for executing the plurality of threads generated by the thread generator, in parallel. Thus, this can provide an appearance inspection apparatus and an appearance inspection method that can execute an appearance inspection at a high speed, irrespectively of a simple configuration.

Abstract: Image processing method comprises providing an original image as a matrix of discreet picture elements, splitting the original image into n frequency channels, each channel being presented by an image matrix of the same size as the original image, detecting edges, and assembling an output (enhanced) image from the n frequency channels, the assembling taking the detected edges into account. The n frequency channels are represented by a low frequency channel and n?1 high frequency channels while splitting the original image into frequency channels, and the edge detection is performed by calculating a correlation value between processed pixel and its neighboring pixels in each of n?1 selected high channels followed by comparing the correlation value with that for the corresponding pixels in other high frequency channels and with the threshold value for this channel.

Abstract: A multi-mode texture compression algorithm is provided for effective compression and decompression texture data during graphics processing. Initially, a request is sent to memory for compressed texture data. Such compressed texture data is then received from the memory in response to the request. At least one of a plurality of compression algorithms associated with the compressed texture data is subsequently identified. Thereafter, the compressed texture data is decompressed in accordance with the identified compression algorithm.

Abstract: A method and system for automated sorting machines is provided in which a feedback between OCR and ABL takes place. According to a first aspect of the present invention, a loose coupling between ABL and OCR is proposed in which the ABL system is asked for a further address block potentially present on the letter's surface when the first one being passed from ABL to OCR yielded that no destination ZIP code could be extracted from it. The confidence of the actual OCR and the ABL result, the address block contents supplemented with some additional criteria based on general knowledge, further, the sort plan actually loaded on the concerned sorting machine and the postmark information found in the stamps of the postmark can advantageously be used to make the ABL restart decision. Further, the ABL system can continue to search further address blocks while the OCR system processes one or a plurality of address blocks being passed to it from ABL in a preceding procedure related to the same letter.

Abstract: Image dividing means of an RPU divides raw image data into divided image data A1 having 2048 horizontal pixels and A2 having 1024 horizontal pixels. The divided image data A1 is continuously processed in single pixel processing means and multiple pixel processing means and thereafter transferred to and stored in a buffer. The divided image data A2 is processed in the single pixel processing means and thereafter transferred to and temporarily stored in another buffer. The multiple pixel processing means reads and processes divided image data A2a stored in this buffer and thereafter transfers and stores the same to and in still another buffer. Image combining means reads divided image data A1b and A2b stored in the buffers and combines the same with each other. Thus, an image processing time and a cost can be reduced even if raw image data having horizontal pixels in a number exceeding the capacity of a line memory is received.

Abstract: All threshold data applied to processing is read into a register so that threshold data read from threshold matrix memory can be recycled until the end of a scanning line for processing, and is selectively output to a plurality of comparison means for executing parallel comparison processing. The threshold data set in the register is shifted in order for repetitive use. The threshold data of the next scanning line is read into the register during the comparison processing, and upon completion of processing of one scanning line, comparison processing on the next scanning line is executed as pipeline processing. Threshold data read from the memory needs to be executed only once for each scanning line for processing, and threshold data can also be read during halftone data generation of the preceding line.

Abstract: The present invention provides a data de-noising system utilizing processors and wavelet denoising techniques. Data is read and displayed in different formats. The data is partitioned into regions and the regions are distributed onto the processors. Communication requirements are determined among the processors according to the wavelet denoising technique and the partitioning of the data. The data is transforming onto different multiresolution levels with the wavelet transform according to the wavelet denoising technique, the communication requirements, and the transformed data containing wavelet coefficients. The denoised data is then transformed into its original reading and displaying data format.

Abstract: A method of and apparatus for processing print data by simultaneously receiving print data from a plurality of host computers, sequentially decomposing the print data, and outputting decomposition results to a print engine. The method includes: storing a decomposition context of a current job when a new urgent job is requested to be processed and the current job is not urgent; decomposing print data for the new urgent job; and resuming decomposition of the interrupted current job using the stored decomposition context.

Abstract: In order to reduce time required to perform error diffusion processing on one line of image data, one line of multi-tone image data 100 is divided into two or more segments of image data 102 and 103 with an overlapped region 101 provided across the boundary. The two or more segments of image data 102 and 103 are subjected to error diffusion to create binary image data 104 and 105 corresponding to the divided image data 102 and 103, respectively. Then the binary image data 104 and 105 are merged to form binary data 106 corresponding to the original image data of one line. Upon merging the binary image data, binary data b2 obtained from the divided image data 102 located upstream of the line before divided in the direction in which the error diffusion processing progresses is used as binary data corresponding to the overlapped region.

Abstract: In order to provide a high-speed, inexpensive inspection system that has a short development period, that is flexible, and that allow algorithms to be easily changed, a PC equipped with an image input feature is used to capture an image detected by a line image sensor, this detected image is transferred to a plurality of PCs connected by a LAN, and defects are detected using software processing on the plurality of PCs.

Abstract: An image processing system includes an image input unit having a plurality of image input channels; a parallel image data processing unit that receives image data from the image input channels of the image input unit, divides the image data received into a plurality of divided data, each divided data including data in a number that is less than a predetermined number, and processes in parallel the divided image data; and a sequential image data processing unit that sequentially processes the divided image data output from the image parallel data processing unit while switching the divided image data for the image input channels.

Abstract: A page to be printed is organized into a plurality of strips. To rendered the page using a tandem printer, each strip is rendered a plurality of times, once for each print engine of the tandem printer. Each time the selected strip is rendered for a different one of the print engines, the rendered image data for that print engine is stored into a separate memory area, such as a buffer, associated with that print engine Once a selected strip is rendered and the rendered image data stored into the associated memory areas, the rendered image data is used by the appropriate print engines to form image portions on a receiving sheet corresponding to the rendered image data. At the same time, a next strip is selected and repeatedly rendered for each of the print engines.

Abstract: When generating and printing an image based on a document written in XHTML or another structured tag language, the throughput of image generation and printing is improved. An apparatus of the present invention comprises judgment means for judging whether a document written in a structured tag language contains tags which can cause disagreement between the order in which the coordinate positions of objects are determined and the order in which object images are formed, and control means to perform control such that, when the above tags are judged to be contained, image formation processing is executed after processing to determine coordinate positions has ended, and when the above tags are judged not to be contained, processing to determine coordinate positions and processing to form images are executed in parallel.

Abstract: An imaging apparatus and method for a single-pass, duplex-capable scanner having increased simplex scanning speed. The imaging apparatus has a first image reading element, a first image processing component, and a second image processing component. When the imaging apparatus is in a simplex mode, a data switch is configured to alternately pass image data acquired through the first image reading element to the first image processing component and the second image processing component.

Abstract: Error diffusion is performed upon input image data. In one aspect, multiple error diffusion processing elements perform error diffusion on a selected pixel in parallel. In another aspect, the error diffusion logic is integrally formed with a fast local memory in the same electronic device, such as an ASIC. The error data produced by the error diffusion logic for a pixel is buffered in the fast local memory until it is to be used by the error diffusion logic on other pixels. In still another aspect, a first-in-first-out (FIFO) buffer regulates or buffers the color image data between the output of a color conversion system, such as a colorant lookup table, and the input an error diffusion processing element. In yet another aspect, the error diffusion logic has tagging logic that produces and stores an indicator, either in the output data stream itself or in a separate area, to indicate whether a raster contains printable data.

Abstract: A printing system comprised of a printer, a plurality of processing nodes, each processing node being disposed for processing a portion of a print job into a printer dependent format, and a processing manager for spooling the print job into selectively sized chunks and assigning the chunks to selected ones of the nodes for parallel processing of the chunks by the processing nodes into the printer dependent format. The chunks are selectively sized from at least one page to an entire size of the print job in accordance with predetermined splitting factors for enhancing printer printing efficiency.

Abstract: A printing system comprised of a printer, a plurality of processing nodes, each processing node being disposed for processing a portion of a print job into a printer dependent format, and a processing manager for spooling the print job into selectively sized chunks and assigning the chunks to selected ones of the nodes for parallel processing of the chunks by the processing nodes into the printer dependent format. The chunks are selectively sized from at least one page to an entire size of the print job in accordance with predetermined splitting factors for enhancing printer printing efficiency. The operating of the printing system includes a method for parallel processing of a print job with a plurality of processing nodes into a printer-ready format for printing the print job, wherein the processing nodes communicate with a virtual disk transfer system and control usage thereof by monitoring available space in the virtual disk transfer system.

Abstract: A scanning or copying system can include imaging elements and one or more system processors that are programmed or adapted to perform image processing methods and algorithms on image data, and in some instances, to enhance the image. Image data is acquired using imaging elements. Some imaging elements may have overlapping or rotated fields of view or employ differing resolutions. For each imaging element, its output is recombined together with the output of one or more other imaging elements. To perform the recombination, the system can extract features in an overlapping region and match these features in multiple images. In some instances, the features matched can be edges. Alternatively, the recombination can be performed by positioning each subimage with respect to a larger image through image matching and location techniques. Parameters from the recombined image can be extracted and these parameters can be used to correct for geometrical and spatial distortions and thereby enhance the image.

Abstract: The invention provides an interpolation processing apparatus, an interpolation processing method and an image display apparatus wherein, when applied, for example, to format conversion for which a programmable video digital signal processor is used, processes such as format conversion can be executed by a comparatively simple process. The programmable video digital signal processor includes a plurality of element processors whose operations are controlled simultaneously and parallelly. The element processors are controlled so as to produce corresponding weighting coefficients based on number data individually allocated thereto. Where image data, for example, for one line are successively inputted to and processed by the element processors, an interpolation process is performed by an arithmetic operation process common to the element processors in which the position information of the number of lines or like is used as a variable.

Abstract: A method and an apparatus for fault-tolerant computing in a parallel processing environment. Intermediate data points are determined from a set of input data by a plurality of processing elements operating in parallel. Preferably, the resolution of the intermediate data points is greater than the resolution required in an output imaging data set. The intermediate data points are spatially re-gridded to the resolution required in the output imaging data set, ignoring any missing data as a result of a processing element failing.

Abstract: A circuit for processing a first image including two image supply blocks, two image processing units, a control unit and a plurality of buses. The image supply blocks assert selected lines of image data onto a respective one of first and second plurality of buses. The image processing units each process the data according to respective algorithms and provide respective update ok signals that each indicate that the respective image processing unit has completed use of the first sub-portion of data. The image supply blocks provide respective update signals to the image processing units in response to the update ok signals from both of the image processing units, transfer data from the second sub-portion to the first, and assert new data on the second sub-portion. Each image processing unit, in response to receiving both update signals, changes state to track the data without losing bus cycles to maintain performance.

Abstract: A symmetric type image filter processing apparatus having a symmetric type image filter composed of symmetric kernel coefficients, in which SIMD commands are utilized efficiently for making the filtering processes high speed, is provided. The symmetric type image filter processing apparatus provides a row-wise intermediate data generating section, a row-wise intermediate data utilizing section, and a memory. The row-wise intermediate data generating section multiplies each kernel coefficient of M pieces in each column of {(N+1)/2} columns at the right or left column by each pixel of M pieces in the column direction of image data having P pixels in one row, and cumulatively adds the multiplied results, by using SIMD commands that can process sequential data of Q pieces. This multiplication and addition operation is executed P/Q times, and intermediate data in one row of the image data are generated and stored in an intermediate data storing region in the memory.

Abstract: Tetrahedral interpolation by rewriting the interpolation in terms of ordered differentials and color differences to lower the computational complexity. Additionally, hardward architecture allows efficient implementation.

Abstract: In order to reduce the total time needed to print data produced by application software, producing of data and printing of the data are performed in parallel. In the parallel processing mode, the application software outputs data part by part in the same order as the order in which the data is printed by a printer on printing paper in a printing direction, that is, in a direction from the top to the bottom of the page. A printer driver declares to an OS (Operating System) that the OS should perform processing in an unbanding mode, and banding is performed by the printer driver. In the parallel processing mode, data of an area is printed when data of a next area is being scanned, and thus the total time from the start of a copying process to the end of printing is reduced.

Abstract: A method and apparatus for multi-threaded, multitasking processing of image data includes inputting a plurality of image data portions representing one or more images to be processed and putting the first portion of the image data in a memory storage portion. The memory storage portions may be called queues, stacks, address spaces, registers, files, arrays and buffers. While receiving a second portion of the image data, a first portion of the image data is analyzed for the need to be modified by one or more processing methods. Then the image data is modified as necessary by the one or more processing methods while possibly receiving additional image data, and analyzing the second portion of the image data for the need to be modified by the one or more processing methods.

Abstract: A method and apparatus for prioritizing the use of multifunctional printing system's basic processing resources for multiple banded image processing. The printing system employs a controller with an improved job contention manager (JCM). A plurality of basic resources of the printing system are provided with a queue. One or more job services, at desired times, signals the JCM to transfer a multiple banded image sub-job of a given job. The signal for each of the bands of the sub-job includes information about the respective sub-job's priority, band number and total band count for the image. Responsive to the signal from the job service the JCM adds a corresponding basic resource sub-job to the queues of each basic resourse which the sub-job will require to perform the sub-job. When the last band of an image has been transferred it is merged into a single image for further processing.

Abstract: First control controls execution of a process for inputting an image drawn by the user using an input window, and a process for computing the image feature amount of the input image. Parallel to the first control, second control controls execution of a process for storing a plurality of image data in correspondence with their image feature amounts, a process for computing image similarity on the basis of the computed image feature amount and those of the stored image data, and a process for displaying a list of image data as search results on the basis of the computed image similarity. Also, communications between the first control and the second control are controlled.

Abstract: A CMOS imager in which a CMOS image sensor, a color image processing module and an image compression module are all provided on a single die. Both the color image processing module and the image compression module incorporate pipelined architectures to process the image data at a video rate in a massively parallel fashion.

Abstract: A system and process for deghosting mosaiced images created by stitching together images of a scene captured from different viewpoints is presented. When images are mosaiced, which were captured by different cameras at different viewpoints, the possibility of localized double images of objects exists. Essentially, this double imaging or ghosting will occur if an object in the scene is close in to the cameras capturing the images. However, this localized ghosting can be compensated for by estimating the amount of local mis-registration and then locally warping each image in the mosaiced image to reduce any ghosting.

Abstract: An image processing apparatus includes an error correcting section for correcting an input image on the basis of a supplied error correction amount and outputting a corrected image, a threshold processing section for processing the corrected image on the basis of a predetermined threshold, further processing the corrected image on the basis of the error correction amount, and outputting an output signal, a correction amount calculating section for calculating in advance a plurality of correction amounts corresponding to possible values of the output signal in parallel with processing of the threshold processing section on the basis of the corrected image output from the error correcting section, and selecting one of the plurality of error correction amounts as the error correction amount in accordance with the value of the output signal actually output from the threshold processing section, and an error correction amount supplying section for supplying the error correction amount calculated and selected by th

Abstract: In a pipeline process structure having a plurality of processing functions 131-1 to 131-N which are connected to each other in a pipeline mode so as to execute predetermined process operations, while an image information read-skipping unit 132 for skipping a reading operation of a portion of image information to be processed is provided with at least one of these plural processing functions 131-1 to 131-N, when the image information is processed on such a system for requiring/releasing a memory, the image information can be effectively processed and acquired.

Abstract: A photoelectric conversion apparatus includes a plurality of elements (D) for receiving reflected light from an object, photoelectrically converting it into image signals (OS1-OS4), and outputting the image signals, a first output buffer (121) for outputting image signals received from even elements of the plurality of elements in the line from one end to the other end, a second output buffer (122) for outputting image signals received from odd elements from one end to the other end, a third output buffer (123) for outputting image signals received from even elements from the other end to the one end, and a fourth output buffer (124) for outputting image signals received from odd elements from the other end to the one end.

Abstract: A system and method for error diffusion screening with parallel processing, in which the calculation of an error diffused output image is divided into two distinct processes. A first process is used to calculate the tone value for an output pixel based only on a corresponding input pixel value. The second process is used to the manage distribution and application of the error terms to output values. The error arising from each pixel is apportioned among its neighbors entirely within the error array. As a result, the two processes run simultaneously, provided that the second process does not start to use the error term for any particular pixel until the first process has complete the calculation of the error term.

Abstract: A resampling technique selects positions for resampled pixels in a resampled tile such that adjacent tiles need not be retrieved to obtain a value for the resampled pixel. Thus, the resampled pixel does not depend on source pixels in adjacent source tiles. As a result, the resampling technique accelerates image resampling by localizing computation of resampled pixels to the source tile that contains the resampled pixel. In an alternate embodiment, the resampling technique combines separately computed partial results for a resampled pixel rather than incurring data transfer and synchronization overhead for transferring whole source tiles.

Abstract: A binary morphological erosion operator image processor which operates on images of any size and dimension with a structuring element of any dimension and shape. The erosion operations are all processed in the 1D domain by first decomposing the image and structuring element into a series of 1D slices, and then further segmenting the structuring element into a plurality of shorter 1D segments. These segments are processed in parallel using an array of parallel processors. Once the erosion operations are completed, the resulting 1D slices are composed into a final eroded image.

Abstract: The efficient use of memory devices is crucial in processing a large amount of data. In particular, the buffer memory is used in an efficient manner by the use of a buffer memory control unit. The efficient use of the buffer memory is implemented in a high-speed color data processing unit in a color image processing device such as copiers, facsimile devices and printers.

Abstract: A method is described for the operation of an opto-electronic sensor array, in which a plurality of sequential operating light pulses are emitted into a monitored region; the operating light pulses reflected from an object arranged in the monitored region are received by a spatially resolving receiver unit with a plurality of light sensitive sensors; the light intensity of the respectively received reflected operating light pulses is detected and stored in each case during an activation interval synchronously with the emission of the operating light pulses for each sensor; and after each activation interval, in sequential transmission steps, the light intensities stored for the sensors are transmitted to a parallel processor having a plurality of parallel signal inputs for the parallel signal processing, with the transmission in each case taking place for a plurality of sensors simultaneously in each transmission step. Furthermore, a sensor array for the carrying out of the method is described.

Abstract: A method and apparatus for prioritizing the use of multifunctional printing system's basic processing resources to permit job streaming. The printing system employs a controller with an improved job contention manager (JCM). A plurality of basic resources of the printing system are provided with a queue. One or more job services, at desired times, signals the JCM to carry out a sub-job of a given job. The signal for each of the sub-jobs includes information about the respective sub-job's, job service and its priority. Responsive to the signal from the job service the JCM adds a corresponding basic resource sub-job to the queues of each basic resource which the sub-job will require to perform the sub-job. A first of the sub-jobs is placed in an “Active” state ready for processing, if the first sub-job is at the top of all of the queues, of all the basic resources, required to perform the first sub-job.

Abstract: In a data conversion device, a plurality of series of data to be converted is input from a buffer memory device to a data processor. The data processor processes a plurality of the series of the data to be converted, simultaneously in parallel, and outputs a plurality of the series of the data to be converted, simultaneously in parallel. A memory controller of the buffer memory device relates a plurality of the series of the data to be converted to respective conversion tables formed therein by the data processor so as to read converted data corresponding to the series of the data to be converted from the respective conversion tables simultaneously in parallel.

Abstract: A linear transform apparatus for implementing a linear transform on input data values to produce linear transformed output data, the apparatus comprising: input means for inputting input data values one after another to each of a series of multiplication means; a series of multiplication means interconnected with the input means for multiplying a current input data value by a constant to produce a current multiplier output; an interconnection network interconnecting the series of multiplication means to predetermined ones of a series of signed accumulator means; a series of signed accumulator means each interconnected to the interconnection network, each of the signed accumulator means producing an intermediate accumulator output by accumulating a corresponding one of the current multiplier outputs with a corresponding previous intermediate accumulator output, each of the signed accumulator means outputting the intermediate accumulator output as a corresponding linear transformed output data value.