Abstract: An integrated circuit for the processing of digitized signals which are representative of a source image which is defined by image elements which are encoded on M bits and which are arranged in I rows and J columns. The image is processed by means of M sliding windows (W.sub.1 to W.sub.M) which consist of N rows and P columns and which step-wise slide past each image element along I rows, but which have been shifted through one row with respect to one another. Processing is performed by bit serial operators which successively act on the bits of each of the N+N-1 image elements. The signals may be serially applied via Q contact pads, where Q=N+M-1, or in parallel via Q contact pads where Q is the first multiple of M which is equal to or larger than N+M-1, the integrated circuit then also comprising a parallel to serial converter circuit.

Abstract: A method and apparatus under software control for pattern recognition utilizes a neural network implementation to recognize two dimensional input images which are sufficiently similar to a database of previously stored two dimensional images. Images are first image processed and subjected to a Fourier transform which yields a power spectrum. An in-class to out-of-class study is performed on a typical collection of images in order to determine the most discriminatory regions of the Fourier transform. A feature vector consisting of the highest order (most discriminatory) magnitude information from the power spectrum of the Fourier transform of the image is formed. Feature vectors are input to a neural network having preferably two hidden layers, input dimensionality of the number of elements in the feature vector and output dimensionality of the number of data elements stored in the database. Unique identifier numbers are preferably stored along with the feature vector.

Abstract: An image motion vector detection apparatus effectively reduces influences of false vectors arising at a certain probability and suppresses time lags. An image fluctuation stabilizer highly stabilizes images or frames without causing unnatural visual feelings. An image motion vector detection apparatus comprises means for detecting motion vectors in each of motion vector detection regions defined in the whole or part of picture frame, means for judging the reliability of the motion vector in each region, and signal processing means which, in obtaining a motion vector of the whole or part of the frame from the motion vectors of the regions and in case the number of motion vectors judged to be reliable is more than or equal to a predetermined number, outputs a representative value of a set of the reliable motion vectors as a motion vector of the frame.

Abstract: An improved ART2 network provides fast and intermediate learning. The network combines analog and binary coding functions. The analog portion encodes the recent past while the binary portion retains the distant past. LTM weights that fall below a threshold remain below threshold at all future times. The suprathreshold LTM weights track a time average of recent input patterns. LTM weight adjustment (update) provides fast commitment and slow recoding. The network incorporates these coding features while achieving an increase in computational efficiency of two to three orders of magnitude over prior analog ART systems.

Abstract: An improved pattern recognition system, using an improved method for merging low-level recognition information with auxiliary contextual information such as a Deterministic Finite Automaton (DFA). The system comprises a low-level shape recognizer for handwriting input, an English Language dictionary organized as a Trie (a special type of DFA), and software to merge the results of the two. An input of digitized handwriting strokes is translated into characters using the shape recognizer and the Trie in tandem, allowing the system to reject nonsense translations at the earliest possible stage of the process and without the overhead traversing the trie from the top with each translation.

Abstract: A method and an apparatus for isolating an area corresponding to a character or word in an optical character recognition device. Original manuscript images are compressed (reduced) in a scanning direction corresponding to the direction of the original manuscript image's line (column), and areas where the black picture elements correpsonding to the compressed original manuscript images are connected with each other are searched by following the contour of the areas of connected black picture elements contained in the compressed images. Then, a minimum coordinate point and a maximum coordinate point are obtained in accordance with the areas determined by searching, and rectangular areas determined by the minimum coordinate point and the maximum coordinate point of the area where the black picture elements are connected with each other are extracted. Finally, each area of the original manuscript images corresponding to the extracted rectangular areas is isolated as an area of image.

Abstract: The apparatus and method of the present invention utilizes three scaling functions for bi-tonal digitized image data such as documents. The digitized document resolution (i.e. from a digitizing scanner) may be different from the resolution of an image output device (i.e. laser printer, CRT display, fax machine, etc.). The invention takes in the scanned image data and scales the data to compensate for the resolution mismatch of the output device. Although the invention is intended for document archive applications (such as scanning documents on microfilm), the scaling can be performed on any bi-tonal image.

Abstract: An image processing method distinguishes an object from within an image area by determining a stable threshold value of light intensity values from within the image area. Received light is converted into image data denoting light intensity values of the received light at a plurality of points from within the image area. A bright portion average value denoting an average of the light intensity values which exceed a predetermined threshold and a dark portion average value denoting an average of the light intensity values which are less than the predetermined threshold are calculated. A new threshold is obtained by applying the bright portion average value and the dark portion average value to a predetermined dividing ratio. The object within the image area is distinguished using the new threshold. The method is also applicable to the use of multivalue thresholds.

Abstract: A computer input system includes a facsimile machine connected to a computer through an interface. A data sheet which is fed to the facsimile machine includes a character pattern row in which a character pattern of data to be entered into the computer is written, and a standard pattern row in which a standard reference pattern corresponding to the character pattern is written. An identification index is marked on the data sheet, and the character pattern, standard pattern, and identification index are transmitted by the facsimile machine from the data sheet to the computer via the interface. The computer includes an arrangement responsive to the character pattern, the standard pattern and the identification index for recognizing and recording the corresponding relationship between the character pattern and the standard pattern.

Abstract: A system for recognition of characters on a medium. The system includes a scanner for scanning a medium such as a page of printed text and graphics and producing a bit-mapped representation of the page. The bit-mapped representation of the page is then stored in a memory means such as the memory of a computer system. A processor processes the bit-mapped image to produce an output comprising coded character representation of the text on the page. The present invention discloses parsing a page to allow for production of the output characters in a logical sequence, a combination of feature detection methods and template matching methods for recognition of characters and a number of methods for feature detection such as use of statistical data and polygon fitting.

Abstract: A filing method and a system for filing image data operate to divide an input image having a dot number larger than a predetermined dot number of a display screen at a fundamental unit equal to the display screen size, to file the divided input images as a plurality of partial image data on an optical disk, and to retrieve the filed images at high speed. The division filing system includes an input image process device for compressing the divided partial image data, and an index generating device, such as a complexity detecting circuit, for extracting either position information of a distinguishable partial image, or address information among the divided partial images. An optical disk device stores either the position information or the address information as well as the compressed image data into an optical disk.

Abstract: Transforms such as the DCT are useful for image compression. One close relative of the DCT is preferred for its arithmetic simplicity. A method is described whereby the image compression is done with no multiplications. Other enhancements are made to improve image quality.

Abstract: A still image processing method which can execute the effective flowing process is provided. In this method, when the flowing process is executed by shifting some or all of the pixels constituting a still image of one image plane in a predetermined direction, if no image information to be shifted exists at the first image position, the image information from elsewhere in the image plane is shifted to the second image position. The image information at the second image position is obtained by shifting the image information in the image plane in the direction opposite to the principal shifting direction.

Abstract: An interpolator for enlarging or reducing a digital image includes an interpolation coefficient memory containing interpolation coefficients representing a one dimension interpolation kernel. A row interpolator receives image pixel values, retrieves interpolation coefficients from the memory, and produces interpolated pixel values by interpolating in a row direction. A column interpolator receives multiple rows of interpolated pixel values from the row interpolator, retrieves interpolation coefficients from the memory, and produces rows of interpolated pixel values by interpolating in a column direction.

Abstract: An image processor employing a plurality of types of scanners including a stored table in which identification numbers allotted to the scanners and scanner control information are registered in pairs, a controller for registering the identification number of a scanner from the operator of the image processor, a controller for requesting the scanner to transmit an identification number representative of its type and receiving such identification number, a controller for setting the registered identification number as the identification number of the scanner when the scanner fails to transmit its identification number within a predetermined amount of time, a reading controller for reading from the stored table scanner control information corresponding to the identification number of the scanner, and a controller to control the scanner in a manner to conform with the scanner control information.

Abstract: A and an apparatus method for quickly and easily identifying the five basic components of leukocyte cells: basophils, eosinophils, neutrophils, monocytes and lymphocytes is disclosed. The method measures the size of each of the cells. Computation of the quantile feature of the cells and their color is then made. BA and EO cells are differentiated from LY and MO and Ne cells using the present method. BA and EO cells can also be differentiated from LY, MO, and Ne cells using methods of the prior art. Once BA and EO cells have been differentiated from the five leukocyte cells, LY cells can be differentiated from MO and NE cells based upon the size of the LY cells. The average color intensity of the LY cells determined is then used as the threshold to differentiate MO cells from NE cells.

Abstract: A colony counting apparatus includes a reading unit with a visual sensor incorporated therein. The presence of a colony is detected at the end of that colony by moving the reading unit along lines selected at an appropriate pitch and by reading data on the individual points set along each line as logical high or low signals. The values obtained in the above-described counting operation are added and stored in an addition/storage device in a CPU, and the results of addition which are stored in the addition/storage device are output to a suitable display by an outputting device when the counting operation is completed.

Abstract: A method for labelling polygons of a geometric layout which includes the steps of scanning a geometric layout during a first scan line pass to detect objects which form a polygon, processing the scan line at each occurrence of an event to detect the objects which contact the scan line, assigning temporary numbers and root designators to the objects which contact the scan line in accordance with a sorting criterion, updating the temporary numbers assigned to the objects to keep the temporary number associated with the earliest root designator of each separately detected polygon, and renaming each object which forms a part of the same polygon with a common label. The step of updating includes the steps of numbering each object in a polygon with a temporary number assigned to a root object of the polygon, and storing, in a sorted order, the root objects which lose their status as root objects during the step of updating.

Abstract: In the present invention a method of compressing a signature signal is disclosed. The signature signal is divided into a plurality of signature segment signals where each segment is encoded by using a modified ring-encoding method, such that the total number of grid points along the perimeter of all the rings can be stored in an 8-bit byte. In one embodiment of the method of the present invention, a Fibonacci series is used to determine the relative spacing of the rings. The present invention also discloses a method and apparatus for electronically processing a POS transaction with a human signature for verification of the transaction, as well as for request of extension of credit by a credit company. The signature is electronically captured, compressed and combined with a transaction signal which is representative of the transaction. The record signal is then processed by the credit company for verification of the transaction or for requesting extension of credit.

Abstract: The present invention efficiently and quickly computes the fit parameters for a noise reduction method wherein a digital image is first decomposed into two types of regions; smooth regions and edge regions. Within the digitized image a plane fit parameter is computed at every point. The computation of the fit parameters incorporates convolutions carried out in an efficient manner by the use of a slide sum technique.