Hewlett-packard Company

Hewlett-packard Development Company, L.P.

Renato Kresch

Neri Merhav

Doron Shaked

Michael Elad

Yacov Hel-Or

A motion vector between a current block and a reference block of a reference frame is determined by calculating the exact linear cross-correlation between the current block and the potential reference blocks of the reference picture. The current block is orthogonally transformed using DCT/DST transforms of a first type without prior zero padding of the current block to generate a current quadruple of transform coefficient blocks. The current quadruple is processed together with a reference quadruple of transform coefficient blocks generated from four of the search region blocks to generate a quadruple of processed transform coefficient blocks. The quadruple of processed transform coefficient blocks is inversely transformed using inverse DCT/DST transforms of a second type to generate a block of exact cross-correlations between the current block and the search region. The motion vector is determined from the block of exact cross-correlations.

Method and apparatus for performing motion estimation in the DCT domain

An error diffusion halftoning system and a method of managing halftoning errors utilize a quantization technique to reduce the required size of a primary error buffer that is needed to diffuse the halftoning errors. By implementing the quantization technique, the primary error buffer can be reduced from an 8-bits-per-bin error buffer to a 2-bits-per-bin error buffer for 256 grayscale. The reduction in bin size decreases the cost of the primary error buffer and, consequently, the cost of an error diffusion halftoning (EDH) device of the system that generates halftone values from grayscale values of a digital image. The quantization technique is executed on cumulated half-toning errors, derived from apportioned halftoning errors associated with previously processed pixels. In addition, the system and method utilize an error diffusion procedure to diffuse quantization errors that are produced from the execution of the quantization technique. In one embodiment, entire quantization errors are sequentially transmitted to a supplemental error diffuser, so that each quantization error can be introduced to the grayscale value of the next pixel to be processed. In another embodiment, the quantization errors are apportioned using predefined multiplicative parameters and eventually diffused to two or more unprocessed pixels that are adjacent to the pixel currently being processed.

System and memory-buffer management method for reducing random access memory requirements in error diffusion halftone applications

A face detection system and a method of pre-filtering an input image for face detection utilize a candidate selector that selects candidate regions of the input image that potentially contains a picture of a human face. The candidate selector operates in conjunction with an associated face detector that verifies whether the candidate regions contain a human face. In the preferred embodiment, the candidate selector includes a linear matched filter and a non-linear filter that operate in series to select the candidate regions from the input image. Initially, the linear matched filter performs a linear correlation on the input image using a filtering kernel to derive a correlation image. The linear matched filter selects regions of the input image that have a local maximum in the correlation image and have correlation values greater than a threshold correlation value. Preferably, the linear correlation is performed in the discrete cosine transform (DCT) domain. The non-linear filter then examines contrast values from various sub-regions of the image regions that were selected by the linear matched filter to screen for suitable candidate regions. The filtering kernel used by the linear matched filter is calculated during a training period, or a non-face detecting period, by a filtering-kernel generator. The filtering kernel is preferably computed utilizing a database of training face images that have been averaged and then masked to remove DC, illumination and noise components.

System and method for face detection using candidate image region selection

A system and a method for classifying input vectors into one of two classes, a target class and a non-target class, utilize iterative rejection stages to first label the input vectors that belong in the non-target class in order to identify the remaining non-labeled input vectors as belonging in the target class. The system and method may be used in a number applications, such as face detection, where the members of the two classes can be represented in a vector form. The operation of the system can be divided into an off-line (training) procedure and an on-line (actual classification) procedure. During the off-line procedure, projection vectors and their corresponding threshold values that will be used during the on-line procedure are computed using a training set of sample non-targets and sample targets. Each projection vector facilitates identification of a large portion of the sample non-targets as belonging in the non-target class for a given set of sample targets and sample non-targets. During the on-line procedure, an input vector is successively projected onto each computed projection vector and compared with a pair of corresponding threshold values to determine whether the input vector is a non-target. If the input vector is not determined to be a non-target during the successive projection and thresholding, the input vector is classified as a target.

Approximated invariant method for pattern detection

An implicit DST-based filter having characteristics defined by a linear convolution kernel that may be causal or noncausal-symmetric. The filter filters an information signal composed of blocks of discrete cosine transform (DCT) coefficients to generate a filtered information signal also composed of blocks of DCT coefficients. The filter comprises multiplying matrices, a deriving module, matrix multiplying modules and a summing module. The multiplying matrices are obtained by absorbing a cosine-to-sine transform and a sine-to-cosine transform into kernel matrices derived from the linear convolution kernel. The deriving module derives intermediate blocks of DCT coefficients from neighboring ones of the blocks of DCT coefficients constituting the information signal. The matrix multiplying modules multiply the intermediate blocks of DCT coefficients by the multiplying matrices. The summing module sums the blocks of DCT coefficients generated by the matrix multiplying modules to generate the blocks of DCT coefficients constituting the filtered information signal.

Implicit DST-based filter operating in the DCT domain

An explicit DST-based filter that comprises a trigonometric transform module, first and second transform coefficient processors (TCPs), an inverse trigonometric transform module and first and second summing arrangements. The trigonometric transform module applies a trigonometric transform to blocks of DCT coefficients related to input blocks of DCT coefficients to generate corresponding input blocks of transform coefficients of a second type ('second coefficients'). The first TCP includes matrix multipliers that generate a multiplied block of DCT coefficients and a multiplied block of second coefficients by multiplying, by diagonal multiplying matrices, intermediate blocks of DCT coefficients derived from the input blocks of DCT coefficients. The second TCP includes matrix multipliers that generate at a multiplied block of DCT coefficients and a multiplied block of second coefficients by multiplying, by diagonal multiplying matrices, intermediate blocks of second coefficients derived from the input blocks of second coefficients. The first summing arrangement sums the multiplied blocks of DCT coefficients to generate a first final block of DCT coefficients, and sums the multiplied blocks of second coefficients to generate a first final block of second coefficients. The inverse trigonometric transform module applies an inverse trigonometric transform to the first final block of second coefficients to generate a second final block of DCT coefficients. The second summing arrangement sums the first and second final blocks of DCT coefficients to generate a block of DCT coefficients constituting a block of a filtered information signal.

Growing community of inventors

Haifa, Israel

Renato Kresch