Abstract: The invention is a method and apparatus for reducing noise in an image. The method and apparatus involves calculating a plurality of directional operators, comparing the directional operators to a predetermined threshold, and applying a filter responsive to the comparing. The method and apparatus computes the directional operators by taking a weighted sum of the absolute differences between a target pixel and its surrounding pixels. The comparison signals to the method or apparatus the existence of a line or edge. If the method or apparatus detects no edge or line, the method applies a smoothing or averaging filter. If the method or apparatus detects an edge or line, the method applies a median filter in the direction with a minimum directional difference.

Abstract: The screened printing data are subjected to a filtering method, the raster information being maintained. Intermediate tonal values are produced by the filtering method, using the binary tonal values, the intermediate tonal values being configured in such a way that, by means of a color space transformation, transformed tonal values and/or intermediate tonal values can be formed which represent a true-to-color proof.

Abstract: A plurality of panels are assembled into a single image. Each of the panels may have different intensities throughout the panel, as well as non-uniformities between panels. The panels are modified using flat-field calibration, panel flattening, and panel connecting techniques. These techniques correct for non-uniformities and provide a cleaner, single image.

Abstract: An digital image input, possibly being either pre-compressed or decompressed, is enhanced; its edges are preserved while any compression artifacts, like blocking and ringing, are simultaneously reduced. The enhancement method enhances images with luminance and chrominance functions, incompletely defined or undefined, on a set of pixels so that the missing information is extrapolated while the image is simultaneously enhanced. The method consists of up to three integrated sub-processes: the image sharpening flow; the de-quantization filtering; and the means of control of the local rate of flow. The image sharpening flow is an iterative nonlinear filtering schema intertwining a local median filter and a suitably chosen linear filter. A local geometric control mechanism allows selective application and adaptation of an algorithm allowing selective removal of local artifacts.

Abstract: A surrounding pixel reference unit reads out and stores image data having a predetermined size from a line buffer. A color judgment unit detects the color of the image (reference image) stored in the surrounding pixel reference unit. A parameter determination unit determines parameters used in an individual noise removal unit on the basis of the color detected by the color judgment unit, and supplies them to the individual noise removal unit. The individual noise removal unit reads out, from the line buffer, an image which is located at the same position as that read out by the surrounding pixel reference unit and has a size according to the parameters supplied from the parameter determination unit, and executes a noise removal process for the readout image in accordance with the parameters supplied from the parameter determination unit.

Abstract: A method is provided for detecting spot defects on an object when an allowable variation (called the “background”) in the appearance of the object can be modeled. Methods are also provided for measuring and classifying detected spot defects. An alignment model is used to align the image of the object, a background model is then used to estimate the (possibly different) background in each region, and each background is substantially removed from the image so as to form a foreground image on which blob analysis can be applied to detect spot defects, the blob analysis using a threshold image that accommodates different noise statistics for each region. The method facilitates robust spot defect inspection of fiber optic end faces, or of any object with different object regions. The method also allows use of blob analysis over a larger range of conditions, including conditions that make simple blob analysis infeasible.

Abstract: A method to detect a structure of interest includes shading a target from ambient light. A first plurality of images is acquired from the target. Background in the first plurality of images is reduced to minimize or eliminate brightness variation in the images and to generate a first plurality background-reduced images. A second plurality of images is acquired from the target. Background in the second plurality of images is reduced to generate a second plurality of background-reduced images. Noise in the second plurality of background-reduced images is reduced to generate a noise-reduced image, which is then multiplied to generate an amplified image. A structure of interest is detected in the amplified image. An imaging device that can be used to carry out the method is also included and includes an illuminator that shades the target from ambient light and scatters light.

Abstract: An impulse-reducing module (200) reduces random noise in video pixels by providing an impulse detector (244) and an impulse reducer, such as a median filter (250). The impulse detector (244) generates filter control signals in response to detection of impulses, and the median filter (250) generates filtered pixel values in response to the filter control signals. The control signals set the median filter to a plurality of filter operating modes.

Abstract: Various components of the present invention are collectively designated as Analysis of Variables Through Analog Representation (AVATAR). AVATAR offers an analog solution to those problems of the analysis of variables which are normally handled by digital means. The invention allows (a) the improved perception of the measurements through geometrical analogies, (b) effective solutions of the existing computational problems of the order statistic methods, and (c) extended applicability of these methods to analysis of variables. The invention employs transformation of discrete or continuous variables into normalized continuous scalar fields, that is, into objects with mathematical properties of density and/or cumulative distribution functions. In addition to dependence on the displacement coordinates (thresholds), these objects can also depend on other parameters, including spatial coordinates. Moreover, this transformation of the measured variables may be implemented with respect to any reference variable.

Abstract: A multi-resolution filter is provided that may be used to attenuate aliasing artifacts in a digital input signal, such as a digital image acquired by a digital camera. The multi-resolution filter reduces the resolution of the digital input signal and filters the reduced resolution signal using a median filter. The output of the median filter is provided to an interpolation filter, which increases the resolution of the median filter's output to produce a digital output signal in which aliasing artifacts from the digital input signal have been attenuated. The multi-resolution filter may be advantageously applied to the chrominance channels of digital images to attenuate aliasing artifacts without undesirably blurring sharp luminance and color boundaries in the digital image. Furthermore, the computational expense associated with the median filter is diminished by reducing the resolution of the digital input signal before filtering it using the median filter.

Abstract: An adaptive intra-field algorithm to de-interlace video data uses two line memories to implement a median filter that can use multiple data points on adjacent lines of display. In a median-mode, the filter outputs the median of the input data to create the value for a location on the intermediate line and can decrease image blurring. But if examination of the two adjacent pixels on two adjacent scan lines is small, as in the case of a vertical image line, the filter operates as a line average filter, to avoid creating image artifacts.

Abstract: The invention is a method and apparatus for reducing noise in an image. The method and apparatus involves calculating a plurality of directional operators, comparing the directional operators to a predetermined threshold, and applying a filter responsive to the comparing. The method and apparatus computes the directional operators by taking a weighted sum of the absolute differences between a target pixel and its surrounding pixels. The comparison signals to the method or apparatus the existence of a line or edge. If the method or apparatus detects no edge or line, the method applies a smoothing or averaging filter. If the method or apparatus detects an edge or line, the method applies a median filter in the direction with a minimum directional difference.

Abstract: A surrounding pixel reference unit reads out and stores image data having a predetermined size from a line buffer. A color judgment unit detects the color of the image (reference image) stored in the surrounding pixel reference unit. A parameter determination unit determines parameters used in an individual noise removal unit on the basis of the color detected by the color judgment unit, and supplies them to the individual noise removal unit. The individual noise removal unit reads out, from the line buffer, an image which is located at the same position as that read out by the surrounding pixel reference unit and has a size according to the parameters supplied from the parameter determination unit, and executes a noise removal process for the readout image in accordance with the parameters supplied from the parameter determination unit.

Abstract: A median filter using an M×N window sorts M×N pixel data along each column in descending order to generate N data strings, each having M sorted pixel data. The N data strings are sequentially stored in a shift register. Data strings in different stages of the shift register are merged with one another in descending order to generate a new sorted data string. Such sorted data strings are again merged with each other to generate a data string having M×(N?1) sorted data. M+1 data surrounding the median value of the data string are merged with a data string having M sorted data in descending order to generate a data string having 2M+1 sorted data. The data value at the median rank of the 2M+1 sorted data is output as the median value of the M×N image data.

Abstract: A noise reduction circuit comprises correlation value calculation means for calculating correlation values in a plurality of directions centered at a pixel to be corrected; correction amount calculation means for calculating, on the basis of the correlation values in the plurality of directions calculated by the correlation value calculation means, a correction amount corresponding to the pixel to be corrected for each of the directions; first correction value calculation means for producing a plurality of candidates for correction values corresponding to the pixel to be corrected on the basis of the correction amount for each of the directions which is calculated by the correction amount calculation means; and second correction value calculation means for calculating the correction value for the pixel to be corrected from the plurality of candidates for correction values corresponding to the pixel to be corrected which are produced by the first correction value calculation means.

Abstract: First similarity values along at least four directions are ascertained within a local area containing a target pixel and weighted averaging is performed by adding the pixel values of pixels around the target pixel value to the pixel value of the target pixel, adding weight along a direction having a small first similarity value (along a direction manifesting a high degree of similarity). By incorporating the pixel value level differences among a plurality of pixels on adjacent lines extending adjacent to the target pixel into the first similarity values, it becomes possible to effectively remove jaggies that are difficult to eliminate in the prior art. Furthermore, by making a judgment on degrees of similarity by incorporating color information such as characteristics differences among different color pixels, a more accurate judgment can be made with regard to the image structure to enable very accurate direction-dependent low-pass filtering.

Abstract: The invention relates to an X-ray examination apparatus and a method for forming an X-ray image by means of a processing unit (2) for the correction of image data. In order to achieve automatic correction of imaging defects that are caused by imperfections in the imaging and processing chain, the processing unit (2) is succeeded by a defect detection unit (3) for the detection of image defects that can be detected on the basis of image parameters that can be extracted from image data acquired during clinical examinations, and is suitable for adapting processing parameters (15–21) that are used in the processing unit (2) in dependence on the detected image defects.

Abstract: The invention relates to a method of processing data, which may be pixels (P[i,j]) representing a sequence of pictures, previously encoded and decoded. The method comprises at least in series a first step (ED) of detecting edge pixels within a picture, followed by a subsequent step (TEST) in which a choice is made from the pixels not detected as edges in the previous step, as to whether these pixels are to be filtered or not. Then, the method comprises a filtering step (SAF) which consists in replacing at least a pixel to be filtered with a pixel belonging to a close neighborhood of said pixel, said close neighborhood comprising said pixel and pixels adjacent to said pixel.

Abstract: A method of deriving from an existing selective image smoothing filter either a corresponding filter for selective image sharpening, or a corresponding filter for both selective image smoothing and selective image sharpening. The selective sharpening filter can be implemented quickly by using implementations of the existing selective smoothing filter and a derived matching non-selective smoothing filter as black boxes and combining their outputs in a simple manner. Alternatively, the derived selective sharpening filter can be implemented by inlining the combination of the implementations of the existing selective smoothing.

Abstract: A method and apparatus for processing image data is described, comprising acquiring a frame of image data and compressing the dynamic range of the frame of image data using a dynamic range compression (DRC) algorithm that utilizes down-sampling, median filtering, and up-sampling. The DRC algorithm comprises down-sampling a frame of image data comprising a first array of pixels to generate a second array of pixels, applying a first median filter to the second array of pixels to generate a blurred array of pixels, up-sampling the blurred array of pixels, and removing at least a portion of low-frequency gradient data generated by previous steps from the frame of image data.

Abstract: A method and apparatus for transforming a web page that contains main content and auxiliary data. The web page is converted into a string containing multiple first values and multiple second values. The first values correspond to formatting code segments within the web page and the second values correspond to text segments within the web page. Further, a low-pass filter is applied to the string containing multiple first values and multiple second values, and the output of the low-pass filter is used to determine the location of the main content within the web page.

Abstract: An impulse-reducing module (200) reduces random noise in video pixels by providing an impulse detector (244) and an impulse reducer, such as a median filter (250). The impulse detector (244) generates filter control signals in response to detection of impulses, and the median filter (250) generates filtered pixel values in response to the filter control signals. The control signals set the median filter to a plurality of filter operating modes.

Abstract: Hardware circuit for median calculation in an active pixel sensor.

Abstract: A method of processing a digital image channel to remove noise, includes the steps of: identifying a pixel of interest; calculating a noise reduced pixel value from a single weighted average of the pixels in a sparsely sampled local region including the pixel of interest; replacing the original value of the pixel of interest with the noise reduced pixel value; and repeating these operations for all of the pixels in the digital image channel.

Abstract: An image processing apparatus and a recording medium which can improve the quality of a color document image. Smoothing conditions which are defined by three parameters of “filter size”, “offset”, and “overlap” are set. An edge image which was generated in an edge detection processing is referred to, and a determination is made as to whether an edge pixel exists in a noticed region. When the edge pixel does not exist in the noticed region, a smoothing processing is carried out. When the edge pixel exists in the noticed region, it is determined inappropriate to effect the smoothing processing, and the process goes to a subsequent processing without effecting the smoothing processing. The processing is carried out until the image scanning ends. When the image scanning ends, the next smoothing conditions are set and the processing which is the same as the one described above is performed.

Abstract: A method for automatic removal of vertical streaks involves a comparison between image data for a pixel and its neighbor pixels to determine whether the pixel is non-homogenous. In a preferred method, the comparison employs an adaptive threshold against which a difference between the pixel and its neighbor pixels is compared, with the adaptive threshold taking into account accumulative information pertaining to pixels generated by a common optical sensor element. A preferred method also includes the step of considering locations of non-homogenous pixels in the image to determine whether the image data associated with the non-homogenous pixels needs to be changed or modified. A preferred method also includes the step of associating the non-homogenous pixels with different image data.

Abstract: Techniques for removing from scanned film regions of “missing data” which may include date regions, or may be the result of spots, scratches or folds on the film. Such techniques are particularly designed to handle larger regions of missing data, such as “thick” scratches. The techniques of the present invention segment missing data regions (which may include characters in a date field) and perform component filtering which involves determining the area/perimeter ratio of each segmented missing data region. Only those regions whose area/perimeter ratio is less than a certain threshold are kept for closest-to-radial-based-function (CRBF) filtering to estimate colors from neighboring pixels to fill in the missing data regions.

Abstract: A method for segmenting video objects in a video sequence that is composed of frames including pixels first assigns a feature vector to each pixel of the video. Next, selected pixels are identified as marker pixels. Pixels adjacent to each marker pixel are assembled into a corresponding a volume of pixels if the distance between the feature vector of the marker pixel and the feature vector of the adjacent pixels is less than a first predetermined threshold. After all pixels have been assembled into volumes, a first score and descriptors are assigned to each volume. At this point, each volume represents a segmented video object. The volumes are then sorted a high-to-low order according to the first scores, and further processed in the high-to-low order. Second scores, dependent on the descriptors of pairs of volumes are determined.

Abstract: A method for post-processing compressed image data for reducing blocking artifacts is suggested, in which a set of primary image data (PID) is decomposed into data sets (IID1, IID2) containing the fine structure of the image with structure being smaller than the blocking artifacts and without the blocking artifacts and structures being larger than or comparable to the blocking artifacts as well as the blocking artifacts, respectively. In the decomposition of the primary image data (PID) a non-linear filtering process (F1) is involved. After decomposition of the primary image data (PID) a filtering process (F3) can be applied to the set of intermediate image data (IID2) containing larger details of the image as well as the blocking artifacts. Therefore, in the process of filtering out the blocking artifacts, the fine structure of the image is not influenced.

Abstract: Noise filtering (3) of a signal (x) is effected by estimating (30) a type of noise in the signal (x) and enabling (30) one of at least two noise filters (310, 311, 312), the enabled noise filter (310,311,312) being a most suitable filter for the estimated type of noise. An approximation of the noise (z) in the signal (x) is obtained by computing (302) a difference between the signal (x) and a noise-filtered (301) version of the signal (x). A kurtosis of the noise is used as a metric for estimating the type of noise. If the estimated type of noise is long-tailed noise, a median filter (312) is enabled to filter the signal. If the estimated type of noise is Gaussian noise or contaminated Gaussian noise, a spatio-temporal filter (310,311) is enabled to filter the signal.

Abstract: Disclosed herein is a filter circuit receiving the values of a pixel to be observed and of each pixel within a window containing the pixel to be observed as input data expressed by integers, for replacing the value of the pixel to be observed with the value of pixel at a predetermined ordinal position in an ascending or descending order within the window, including: an extreme value retaining means for obtaining and retaining an extreme value of the values of the pixels within the window; and a data retrieval determination means receiving extreme value data from the extreme value retaining means and the input data, for obtaining the value at a predetermined ordinal position in the ascending or descending order by cumulatively adding frequencies of occurrence of the values of the pixels within the window.

Abstract: The invention concerns a method for comparing two recorded pixelated source images (2a, 2b) of the same rectangular format representing in contrasted form the equipotential lines on an integrated circuit chip. The method consists in subjecting each source image (2a, 2b) to an adaptive thresholding processing with three contrast levels, then forming a result image (14) from the processed source images (3a, 3b) by assigning to each pixel a contrast level defined according to a logical rule of comparison.

Abstract: A method for determining a median value in a sliding window environment is provided. The method includes creating a merge tree defining leaf nodes associated with columns corresponding to a first position of a sliding window. A display controller and a computer readable medium having program instructions for determining a median value in a sliding window environment are also provided.

Abstract: The present invention relates to an interpolation method for enlarging a digital image or predicting a moving vector of a compressed image system as a sub-pixel unit when the image digitized through a CCD (Charge Coupled Device) camera ect. has a low resolution in a video phone or video conference or general digital video system, particularly the present invention can be adapted to a post processor of a compressed digital image in order to improve the image quality, and can be used for finding a moving vector of a moving picture compressed type, accordingly the present invention is capable of improving the image quality.

Abstract: Feature quantities of a transmitted picture and a received picture are extracted from a first and second block-by-block feature quantity extraction parts which are provided on a transmitting and receiving sides, respectively. The extracted feature quantities are transmitted to a block degradation calculation part which compares the transmitted feature quantities and finds the degree of picture quality degradation for each block. Next, a median filter replaces the degree of picture quality degradation for the each block with a median of the degrees of picture quality degradation among the neighboring blocks of the each block. Then, a degraded block detection part compares the median with a predetermined threshold and detects a degraded block. Finally, a degraded region detection part removes an isolated degraded block and detects a degraded region. According to the invention, local picture quality degradation in a frame caused by transmission failure can be automatically detected with high accuracy.

Abstract: A method and system of noise filtering is provided. The pixels of a first filter mask are separated into groups based on luminance . . . The sizes of each group is determined and a largest group is selected. The distance of each group of pixels from the largest group is also calculated. Pixels in groups that are small compared to the largest group and far from the largest group are tagged as noisy. After tagging the noisy pixels, additional filtering can be applied to the pixels of first filter mask without degradation from the tagged pixels.

Abstract: The present invention includes a method of computing a function array in reconfigurable hardware that includes forming in the reconfigurable hardware a first delay queue and a second delay queue, inputting from a source array outside the reconfigurable hardware a first value into the first delay queue and a second value into the second delay queue, defining in the reconfigurable hardware a window array comprising a first cell and a second cell, inputting the first value from the first delay queue into the first cell and the second value from the second delay queue into the second cell, and calculating an output value for the function array based on the window array. The present invention also includes a method of loop stripmining and a method of calculating output values in a fused producer/consumer loop structure.

Abstract: An adaptive median filter (40) provides dynamic detection and correction of digital image defects which are caused by defective or malfunctioning elements of a radiation detector array (20). The adaptive median filter receives (100) lines of pixel values of a digital image that may have defects and a user-defined defect threshold. The lines of pixel values are scanned on a pixel-by-pixel basis using a kernel of n×n pixels, where the kernel contains the candidate pixel being examined (120). Each kernel is numerically reordered (130) and a median value is calculated (140). A defect threshold value is calculated by multiplying the user-defined defect threshold criteria and the candidate pixel value (150). A reference value is calculated by subtracting the candidate pixel value and the median value (160). The reference value is compared to the defect threshold value (170).

Abstract: An image filter circuit which realizes a large-scale median filter as a digital circuit which has not been conventionally constructed without difficulty. The filter circuit has a histogram A update unit 102 which generates a histogram A with upper-bit data of pixels within a matrix, a cumulative adder A104 which searches for a class including a pixel value corresponding to a predetermined rank in ascending order as a target class based on the histogram A, a histogram B update unit 106 which generates a histogram B with lower-bit data of pixels included in the target class, and a cumulative adder B108 which searches for a pixel value of the predetermined rank in ascending order in the matrix. The filter circuit having these constituent elements outputs a median pixel value of the matrix.

Abstract: Abnormal regions in volumetric image sets are detected and delineated through the following technique. Noise is suppressed in the original data. The background is classified into one or more background classes. An exemplar is identified. Essentially similar structures throughout the volume are identified; a directed clustering technique has been developed for doing so. Quantitative information (lesion volume, shape, etc.) is extracted and output to database.

Abstract: For the purpose of conducting three-dimensional spatial filtering with coefficients dependent upon a pixel value throughout a three-dimensional image, in conducting filtering on the three-dimensional image by a three-dimensional spatial filter, coefficients of the three-dimensional spatial filter are changed according to a property of a pixel value of the three-dimensional image. The property of the pixel value is represented by the pixel-of-interest value, the average pixel value, maximum pixel value, minimum pixel value, or median pixel value of the pixel-of-interest and its neighboring pixels, or the property of the pixel value, the sum of absolute values or the square sum of the differences between pixel values of the pixel-of-interest value and the other pixels.

Abstract: A system to minimize the number of comparison and rearrangement operations and increase the processing speed with a small circuit scale utilizing a median filter in the form of hardware. A signal processing apparatus for performing, when a region to be subjected to signal processing has shifted, signal processing of pixel signals in a region to be subjected to signal processing after shifting the region, by utilizing old pixel signals already subjected to signal processing before the shift of the region. The apparatus includes a new-data-order determination circuit for determining an order of magnitude of pixel signals newly added to a region to be subjected signal processing after the shift of the region. A new-data selection circuit rearranges the new pixel signals based on the determined order. An all-data-order determination circuit determines the order of magnitude of all signals by comparing the rearranged new pixel signals with the old pixel signals.

Abstract: The present invention includes data imaging methods and image forming devices. One aspect of the present invention provides a data imaging method including the steps of providing an image forming device including a filter interface; communicating at least one data stream within the image forming device; adaptively interfacing a filter with the filter interface of the image forming device; accessing at least one data stream using the filter following the interfacing; and forming an image following the accessing.

Abstract: A surrounding pixel reference unit reads out and stores image data having a predetermined size from a line buffer. A color judgment unit detects the color of the image (reference image) stored in the surrounding pixel reference unit. A parameter determination unit determines parameters used in an individual noise removal unit on the basis of the color detected by the color judgment unit, and supplies them to the individual noise removal unit. The individual noise removal unit reads out, from the line buffer, an image which is located at the same position as that read out by the surrounding pixel reference unit and has a size according to the parameters supplied from the parameter determination unit, and executes a noise removal process for the readout image in accordance with the parameters supplied from the parameter determination unit.

Abstract: An digital image input, possibly being either pre-compressed or decompressed, is enhanced; its edges are preserved while any compression artifacts, like blocking and ringing, are simultaneously reduced. The enhancement method enhances images with luminance and chrominance functions, incompletely defined or undefined, on a set of pixels so that the missing information is extrapolated while the image is simultaneously enhanced. The method consists of up to three integrated sub-processes: the image sharpening flow; the de-quantization filtering; and the means of control of the local rate of flow. The image sharpening flow is an iterative nonlinear filtering schema intertwining a local median filter and a suitably chosen linear filter. A local geometric control mechanism allows selective application and adaptation of an algorithm allowing selective removal of local artifacts.

Abstract: Embodiments of a method and/or apparatus for retrieving images are disclosed.

Abstract: A method and system for performing coronary artery calcification scoring using an imaging system including obtaining image data; examining the image data so as to identify a plurality of discrete pixel elements, wherein each of the pixel elements includes a pixel value; dividing the pixel elements into a scorable region so as to create a plurality of scorable pixel elements; processing the scorable pixel elements so as to identify a center scorable pixel element and determine a median pixel value; and replacing the pixel value for the center scorable pixel element with the median pixel value. Also claimed is a medium encoded with a machine-readable computer program code for performing coronary artery calcification scoring, the medium including instructions for causing controller to implement the aforementioned method.

Abstract: The present invention concerns a method and an arrangement for reducing noise in a substance raw data image containing noise. The method may comprise following steps in any order for removing different noise signals contributions step-by-step in any order in a substance raw data image: reducing background radiation; reducing peak disturbance; locating the detection area within the search area; moving binary artifacts; removing unwanted areas of the detection area; and applying default detection area in noisy images). The present invention also relates to a method and an arrangement for determining a measure of at least one substance.

Abstract: A noise reduction system comprises an inputting unit inputting image data, and a noise reduction unit performing a process for removing noise of the image data input by the inputting unit. The noise reduction unit includes a representative value calculating unit calculating a representative value from the level values of a plurality of pixels in a predetermined direction, which include an observed pixel of the input image data, a representative value selecting unit selecting one representative value according to a predetermined condition from among a plurality of representative values which are calculated by the representative value calculating unit and correspond to a plurality of directions, and a replacing unit replacing the level value of the observed pixel with the representative value selected by the representative value selecting unit.

Abstract: A median filter using an M×N window sorts M×N pixel data along each column in descending order to generate N data strings, each having M sorted pixel data. The N data strings are sequentially stored in a shift register. Data strings in different stages of the shift register are merged with one another in descending order to generate a new sorted data string. Such sorted data strings are again merged with each other to generate a data string having M×(N−1) sorted data. M+1 data surrounding the median value of the data string are merged with a data string having M sorted data in descending order to generate a data string having 2M+1 sorted data. The data value at the median rank of the 2M+1 sorted data is output as the median value of the M×N image data.