Abstract: A system and method are disclosed for automatically classifying images of pages of a source, such as a book, into classifications such as front cover, copyright page, table of contents, text, index, etc. In one embodiment, three phases are provided in the classification process. During a first phase of the classification process, a first classifier may be used to determine a preliminary classification of a page image based on single-page criteria. During a second phase of the classification process, a second classifier may be used to determine a final classification for the page image based on multiple-page and/or global criteria. During an optional third phase of classification, a verifier may be used to verify the final classification of the page image based on verification criteria. If automatic classification fails, the page image may be passed on to a human operator for manual classification.

Abstract: This invention provides a novel single-pass and multi-pass synchronized encoder and decoder, performing order(s) of magnitude faster data compression and decompression, at any compression ratio with the higher or the same perceived and measured decompressed image quality in comparison with the best state-of-the-art compression methods, using order(s) of magnitude less system resources (processor complexity, memory size, consumed power, bus bandwith, data latency). These features are achieved using novel direct and inverse non-stationary filters for the recusive octave direct and inverse subband transformation, novel simple context modeling and symbol probability estimation using a minimum number of histograms with the fast adaptation for the sign and the magnitude of the transformation coefficients, a novel accelerated range coder without division operations, and a novel synchronisation of the compressed data.

Abstract: An information source encoding method for encoding a Gaussian integer signal includes the steps of: inputting a signal value sequence of a Gaussian integer signal as an encoding target; transforming signal values included in the input signal value sequence into integer pairs, each having two integers, arranged in the input order; regarding each of the integer pairs as a lattice point on two-dimensional coordinates, and obtaining integer values greater than or equal to zero by performing a two-dimensional-to-one-dimensional mapping in which the shorter the distance from each lattice point to the origin, the smaller the value assigned to the lattice point by the mapping; and encoding the integer values using codes which are used for encoding an information source that follows an exponential distribution.

Abstract: A reflectance image that represents a reflectance distribution, and an illuminance component image that represents an illuminance distribution are generated from an input image. A plurality of small regions, which are divided based on illuminance components of the generated illuminance component image, are specified. A quantized image is generated from the reflectance image for respective specified small regions. Regions having equal quantized pixel values are acquired in the quantized image. Representative pixel values for the respective acquired regions are acquired based on the illuminance component image. Quantized pixel values of the respective acquired regions are corrected using the acquired representative pixel values.

Abstract: A method for performing a color based white balance of a digital image, the method includes: generating or receiving multiple chromaticity value pairs of multiple pixel areas of the digital image; finding selected candidates; wherein a selected candidate represents a pixel area and has a chromaticity value pair that falls within at least one color gamut out of a group of color gamut of different colors; wherein the different colors comprise at least one color that differs from gray; calculating at least one white balance compensation parameter in response to pixel information of pixels areas that have selected chromaticity values; and providing a compensated digital image by applying at least one white balance compensated image.

Abstract: Embodiments of the present invention comprise systems and methods for refining text-detection results for a digital image.

Abstract: An image processing device includes: a saturated area extraction unit that extracts a saturated area corresponding to a saturated pixel from an image expressed by a color component signal of a predetermined color system corresponding to a pixel; a color information acquisition unit that acquires color information from a color component signal corresponding to a pixel contained in an area in contact with the saturated area of the image; and a color component signal modification unit that modifies a color component signal of a pixel in the saturated area so that color information on the saturated pixel contained in the saturated area approaches the color information acquired by the color information acquisition unit.

Abstract: What is disclosed is a novel system and method for performing spatial gamut mapping on a received input color image having a plurality of pixels. A standard gamut-mapping algorithm is applied to the input color image to produce a gamut-mapped color image. A difference is computed between a selected channel of the input color image and the gamut-mapped image to produce a difference image. A local measure of complexity is derived for a given pixel in the difference image. One or more parameter values of a spatial bilateral filter are obtained from a lookup table based on the computed local measure of complexity. The spatial bilateral filter is applied, using the obtained parameter values, to the current pixel of the difference image to produce a modified pixel in a modified difference image. Thereafter, a modified gamut-mapped color image is obtained from the modified difference image and the gamut-mapped color image.

Abstract: This invention provides a technique of easily encoding image data to generate encoded data having high image quality within a target code amount using a small memory capacity by image encoding processing of performing frequency transform and quantization of each pixel block. A frequency transform unit separates image data into low frequency band data and high frequency band data. A coefficient quantizing unit, coefficient encoder, and code amount controller operate to encode the high frequency band data within a predetermined amount. When the encoding processing of the high frequency band data has ended, the quantization parameter of the low frequency band data is set based on the generated code amount of the high frequency band data. A coefficient quantizing unit, coefficient encoder, code amount detector, and quantization parameter updating unit operate to encode the low frequency band data into codes within a low frequency band target code amount.

Abstract: The present invention relates to an encoding apparatus, an encoding method, an encoding program, a decoding apparatus, a decoding method and a decoding program for encoding and decoding efficiently a video signal in interlace scanning. An adder 12 and a subtractor 13 respectively perform addition and subtraction operations on an odd field and an even field of moving image data in interlace scanning. An intra encoder 14 and an intra encoder 15 wavelet transform addition image data and subtraction image data, quantize transform coefficients W, and entropy encode quantization coefficients q. Encoded data output from the intra encoder 14 and the intra encoder 15 is multiplexed into a single encoded stream to be output. Since data with the addition and subtraction operations performed on the odd field and the even field thereof is encoded, dynamic range is doubled while compression efficiency is increased.

Abstract: An apparatus and method for preparing a stain to match a target color of a selected substrate (e.g., wood species). The apparatus and method may include a computer system and hardware or software to display to a user of such system an image array depicting a target image of a stained building material and a plurality of images of colored stains as applied to the same or other building materials.

Abstract: An image noise reduction method is provided. An image is received. A first-stage process is performed to the image to obtain a luminance information Y and a color information Cb and/or Cr corresponding to a pixel array in an YCbCr domain. A second-stage process is performed to the luminance information Y to reduce at least a luminance noise. A third-stage process is performed to the color information Cb and/or Cr to reduce at least a color noise. The luminance information Y and the color information Cb and/or Cr are then combined.

Abstract: Aspects of the subject matter described herein relate to image restoration for compressed images. In aspects, image restoration is accomplished by recovering spectral information from data corresponding to a compressed image. The spectral information is recovered using an algorithm to search through a solution space of possible solutions while constraints are imposed on the solution space to trim undesirable solutions from the space. An algorithm described herein may be iteratively applied to improve the quality of the recovered image.

Abstract: A compensation method for alleviating color shading in a digital image is adapted to correct a color shading phenomenon in a digital image that causes luminance differences between regions in the digital image.

Abstract: The present invention relates generally to pre-processing images, audio and/or video to improve biometric analysis from such. In one implementation, a method is provided including: receiving a color digital image including a representation of a human subject; converting the color digital image into grayscale values; transforming at least one representation of the converted color image; analyzing the transformed, converted color image to identify artifacts; if artifacts are found, processing the color digital image to reduce the artifacts; and providing the processed digital image to a biometric system. Other implementations are provided as well.

Abstract: A method for adjusting a skin color of a digital image adjusts the skin color of an input image. The method includes performing a skin color detection process on the input image to generate a skin-color probability plot Sp in a size corresponding to the input image; providing a hue-saturation lookup table named LUT_Color; performing a skin-color reproduction process on the input image to look up the LUT_Color for a chrominance pixel value for each pixel value of the input image to generate a first image, and adjust each pixel value of the first image by using the skin-color probability plot Sp to generate a second image; performing a skin color smoothing process on the second image to generate a third image; and mixing pixel values of the input image and the third image to generate a target image.

Abstract: In this invention, image data expressed by one component is encoded at a high speed by using a color image lossless encoder. To do this, a color conversion unit converts color image data read by a document reading unit into C, M, Y, and K data. In a color reading mode, a switching unit directly outputs the C, M, Y, and K data to an encoding unit. If the reading mode is a monochrome reading mode, the switching unit neglects the C, M, and Y components of the C, M, Y, and K data. Every time four K components are input, the switching unit supplies the four K components to the encoding unit as pseudo data of C, M, Y, and K color components. The encoding unit lossless-encodes the received C, M, Y, and K component data.

Abstract: Input image data is converted into vector data. The type of the input image data is determined. If it is determined that the input image data is of a first data type, information usable for a search is extracted from the input image data in processing of converting the input image data into the vector data, and information usable for a search is further extracted from the vector data later in the idle time of the image processing apparatus. If it is determined that the input image data is of a second data type, information usable for a search is extracted by performing region segmentation of the input image data. The extracted information is held in association with the vector data as additional information.

Abstract: A color management module which provides color values in a destination color space by interpolation of a LUT that maps from color values in a source color space to corresponding color values in the destination color space. The LUT includes cells corresponding to color values within a spectrum locus and color values outside the spectrum locus. The LUT is populated differently for cells within the spectrum locus and for those outside the spectrum locus. For cells within the spectrum locus, color values are calculated using a color transform constructed based on device profiles for the source device and for the destination device, and corresponding cells of the LUT are populated based on the calculated values. For cells outside of the spectrum locus, an artificial neural network is trained using the calculated color values, and the corresponding cells are populated based on outputs of the trained neural network.

Abstract: A method for performing a color gamut based white balancing of a digital image, the method includes: determining at least one filtering parameter based upon a relationship between candidates and color gamuts that are associated with different illumination conditions; wherein each candidate represents a pixel area; selecting selected candidates by filtering the candidates; wherein the filtering is responsive to at least one filtering parameter; calculating at least one white balance compensation parameter in response to pixel information of the selected pixel areas; and providing a compensated digital image by applying at least one white balance compensated parameter.