75 results about "Scalar quantization" patented technology

An apparatus and method for image data compression performs a modified zero-tree coding on a range of image bit plane values from the largest to a defined smaller value, and a vector quantizer codes the remaining values and lossless coding is performed on the results of the two coding steps. The defined smaller value can be adjusted iteratively to meet a preselected compressed image size criterion or to meet a predefined level of image quality, as determined by any suitable metric. If the image to be compressed is in RGB color space, the apparatus converts the RGB image to a less redundant color space before commencing further processing.

A multimedia compression system for generating frame rate scaleable data in the case of universally scaleable data. Universally scaleable data is scaleable across all of the relevant characteristics of the data (e.g., frame rate, resolution, and quality for video). The scaleable data generated by the compression system includes multiple additive layers for each characteristic across which the data is scaleable. For video, the frame rate layers are additive temporal layers, the resolution layers are additive base and enhancement layers, and the quality layers are additive index planes of embedded codes. Various techniques can be used for generating these layers (e.g., Laplacian pyramid decomposition or wavelet decomposition for generating the resolution layers; tree structured vector quantization or tree structured scalar quantization for generating the quality layers). The system further provides for embedded inter-frame compression in the context of frame rate scalability, and non-redundant layered multicast network delivery of the scaleable data.

A video encoder system includes a base layer and an enhancement layer for encoding video data. The base layer encodes a reduced quality version of the video data to obtain base layer data. The enhancement layer encodes the video data using energy-concentrating transform operations, nested scalar quantization, and Raptor encoders. The base layer data and enhancement layer data are transmitted through a channel to a video decoder system. The decoder system decodes the base layer data to recover an estimate of the reduced quality video and decodes the enhancement layer data (using the reduced quality video as side information) to obtain blocks of coset indices. The decoder system then operates on the blocks of coset indices to generate estimates of the original video data.

Systems and methods that implement compress-forward (CF) coding with N-PSK modulation for the relay channel are disclosed, where N is greater than or equal to two. In the CF scheme, Wyner-Ziv coding is applied at the relay to exploit the joint statistics between signals at the relay and the destination. Quantizer design and selection of channel code parameters are discussed. Low-density parity check (LDPC) codes are used for error protection at the source, and nested scalar quantization (NSQ) and irregular repeat accumulate (IRA) codes for Wyner Ziv coding (or more precisely, distributed joint source-channel coding) at the relay. The destination system decodes original message information using (a) a first signal received from the source in a first interval and (b) a second signal that represents a mixture of transmissions from the source and relay in the second interval.

Disclosed are scalable quantizers for audio and other signals characterized by a non-uniform, perception-based distortion metric, that operate in a common companded domain which includes both the base-layer and one or more enhancement-layers. The common companded domain is designed to permit use of the same unweighted MSE metric for optimal quantization parameter selection in multiple layers, exploiting the statistical dependence of the enhancement-layer signal on the quantization parameters used in the preceding layer. One embodiment features an asymptotically optimal entropy coded uniform scalar quantizer. Another embodiment is an improved bit rate scalable multi-layer Advanced Audio Coder (AAC) which extends the scalability of the asymptotically optimal entropy coded uniform scalar quantizer to systems with non-uniform base-layer quantization, selecting the enhancement-layer quantization methodology to be used in a particular band based on the preceding layer quantization coefficients. In the important case that the source is well modeled as Laplacian, the optimal conditional quantizer is implementable by only two distinct switchable quantizers depending on whether or not the previous quantizer identified the band in question as a so-called "zero dead-zone:" Hence, major savings in bit rate are recouped at virtually no additional computational cost. For example, the proposed four layer scalable coder consisting of 16 kbps layers achieves performance close to a 60 kbps non-scalable coder on the standard test database of 44.1 kHz audio.

The invention discloses a channel information feedback method and a system. The channel information feedback method comprises the following steps: a pre-coding matrix required to be fed back can be decomposed into the product of a plurality of Givens rotating matrixes, wherein each rotating matrix is only related to a rotating angle; and the corresponding rotating angle of the rotating matrix after being decomposed is fed back. The method and the system provided by the invention are a channel compression quantized feedback solution based on the Givens transformation, wherein the pre-coding matrix required to be fed back can be transformed into a limited angle value through continuous Givens rotation, and a base station can rebuild a channel matrix by using the fed-back quantitative angles so as to calculate a pre-coding vector, so that the feedback cost can be effectively reduced. Compared with the traditional scalar quantization, the channel information feedback method can effectively reduce the number of the quantitative elements, thereby decreasing the feedback quantity. With respect to the vector quantization, the only requirement is to carry out the operation of decomposition and quantization, so that the system complexity can be effectively lowered under the condition of ensuring the system performance.

The invention relates to a method of quantizing coding and decoding and a device. The method mainly comprises: on quantizing coding end, if the nearest adjoint point value in a lattice vector is not in a basic code book, a processing mode of splitting the nearest addjoint point value is adopted to process a coding operation; corresponding, on quantizing decoding end, a correspnding processing mode is adopted to process a decoding operation. The invention embodiment can effectively improve processing efficiency in a course of quantizing coding and decoding via combining a basic code book coding mode and a scalar quantization coding mode.

The invention provides a method and apparatus for neural network quantization. Apparatuses and methods of manufacturing same, systems, and methods for performing network parameter quantization in deepneural networks are described. In one aspect, diagonals of a second-order partial derivative matrix (a Hessian matrix) of a loss function of network parameters of a neural network are determined andthen used to weight (Hessian-weighting) the network parameters as part of quantizing the network parameters. In another aspect, the neural network is trained using first and second moment estimates ofgradients of the network parameters and then the second moment estimates are used to weight the network parameters as part of quantizing the network parameters. In yet another aspect, network parameter quantization is performed by using an entropy-constrained scalar quantization (ECSQ) iterative algorithm. In yet another aspect, network parameter quantization is performed by quantizing the network parameters of all layers of a deep neural network together at once.

A system and method for video coding. Energy-concentrating transform operations are performed on video data to obtain transformed data. Nested scalar quantization is performed on the transformed data to generate blocks of coset indices. Bit planes of the blocks of coset indices are encoded using irregular repeat accumulate (IRA) encoders to generate corresponding bit streams. The bit streams are transmitted to a destination device over a channel. A decoder of the destination device receives input data, corresponding to transmitted bit streams, from the channel. The input data is decoded, using side information, to obtain estimates for the blocks of coset indices. Output video data (i.e., an estimate of the original video data) is generated using the estimated blocks of coset indices and the side information.

Provided is a method for constructing an image database for object recognition, which includes a feature extraction step of extracting local descriptors from object images which are to be stored in an image database, a scalar quantization step of quantizing a numeric value indicating each dimension of each of the local descriptors into a predetermined number of bit digits, and a storing step of organizing each of the local descriptors after the quantization to be able to be searched for in the closest vicinity, giving to the local descriptor an identifier of the image from which the local descriptor has been extracted, and storing the local descriptor to which the identifiers are given in the image database. The storing step comprises extracting the local descriptors from the object images when a search query is given, scalar-quantizing each dimension, determining a local descriptor in the closest vicinity of each of the local descriptors from the image database, and storing each local descriptors so as to be able to identify one image by majority vote processing from the images including any determined local descriptor. The scalar quantization step comprises quantizing each dimension of each of the local descriptors into 8 bits or less. Also provided are a processing program for the method and a processing device for performing the processing.

The system carries out conversion of digital video signals organized in blocks of pixels from a first format to a second format. The second format is a format compressed via vector quantization. The vector quantization is performed by means of repeated application of a scalar quantizer to the pixels of said blocks with a quantization step (Q) determined in an adaptive way according to the characteristics of sharpness and/or brightness of the pixels.

Systems and methodologies are described that facilitate generating and/or utilizing explicit and implicit feedback related to a forward link channel for linear precoding in a time division duplex (TDD) multiple-input multiple-output (MIMO) system. Implicit feedback may be provided by estimating a reverse link channel, which may be substantially similar to at least a portion of the forward link channel (e.g., based upon reciprocity). Moreover, explicit feedback may be yielded by quantizing at least part of an estimate of the forward link channel (e.g., utilizing vector and/or scalar quantization).

The system carries out conversion of digital video signals organized in blocks of pixels from a first format to a second format. The second format is a format compressed via vector quantization. The vector quantization is performed by means of repeated application of a scalar quantizer to the pixels of said blocks with a quantization step (Q) determined in an adaptive way according to the characteristics of sharpness and/or brightness of the pixels and representing said vector quantization in a n-dimensional space indicative of the characteristics on n of said pixels in the block partitioned into cells of size proportional to said quantization step, each cell being assigned to an appropriate binary code, wherein said process further includes identifying at least one symmetry element in said n-dimensional space suitable for separating at least two symmetrical set of cells, and selecting one of said at least two symmetrical set of cells for the assignment of said binary codes. A symmetrical permutation on the n pixels of the block is performed according the selection and a part of said binary code indicating the status of said symmetrical permutation is conveniently set.

The invention relates to a self-adaptive compression method achieving matching of joint photographic experts group (JPEG) image size upper limits of an intelligent transportation system. The method includes: acquiring parameter information of a source image, calculating an integrated compression quality factor, calculating weighting factors of compression parameters, re-sampling for the source image according to the need, calculating an integrated quantization step factor and revising a standard quantization table to a special quantization table, carrying out ranking and setting up different JPEG coding options, carrying out discrete cosine transformation, scalar quantization processing and Huffman coding to obtain JEPG image information actually input, and storing the information into a self-learning parameter bank. The self-adaptive compression method achieving the matching of the JPEG image size upper limits of the intelligent transportation system achieves automatic compression processing with the upper limits capable of being matched with high-definition JPEG vehicle images to facilitate data storage and transmission, and achieves the purpose of effectively controlling the size and the quality of the images according to the need. The usage is flexible and convenient, the working performance is stable and reliable, and the range of application is relatively wide.

The application discloses a scalar quantization using bit-stealing for video processing and its application in video coding. Systems, methods, and apparatus for adaptively zeroing out transform coefficients utilizing a bit-stealing parameter are presented herein. A partitioning component can be configured to separate an image into blocks of video data. Further, a frequency transform component canbe configured to transform pixels of a block of the blocks into transform coefficients. Furthermore, a bit-stealing quantization component configured to predefine quantization intervals. Moreover, the bit-stealing quantization component can modify a size of a quantization interval of the quantization intervals based on a variable bit-stealing parameter. Further, the bit-stealing component can quantize a transform coefficient of the transform coefficients, based on the quantization interval, to a quantized coefficient value of quantized coefficient values.

The invention discloses an EGT-and MRC-based phase TCQ (trellis coded quantization) method for an MISO (multiple input single output) wireless system. The method is characterized by including the steps of firstly, a sending end sending beam forming by EGT, and a receiving end receiving the sent beam forming; secondly, quantifying a vector phase of beam forming by an n-bit TCQ quantizer; and thirdly, the receiving end feeding back quantization result to the sending end. The existing phase quantization technology includes SQ (scalar quantization) and VQ (vector quantization), an SQ method is simple but low in flexibility and performance, the VQ is higher in flexibility and performance than the SQ, the sending end and the receiving end require codebook storage, and design of high-dimensional VQ is difficult. The phase is quantified by TCQ, and the phase TCQ method is high in performance without codebook storage and is easy to implement.

A method of (and concomitant computer software embodied on a non-transitory computer-readable medium for) generating speech comprising receiving a mel-frequency cepstrum employing a set of weighting functions, generating a pseudo-inverse of the set, reconstructing a speech waveform from the cepstrum and the pseudo-inverse, and outputting sound corresponding to the waveform. Also a corresponding method of (and concomitant computer software embodied on a non-transitory computer-readable medium for) encoding speech comprising receiving sounds comprising speech, computing mel-frequency cepstral coefficients from the sounds using a quantization method selected from the group consisting of non-uniform scalar quantization and vector quantization, and generating and storing codewords from the coefficients that permit recreation of the sounds.