39 results about "Quantization distortion" patented technology

A FAWNA that allows high-speed mobile connectivity by leveraging the speed of optical networks. Specifically, SIMO FAWNA, which comprises a SIMO wireless channel interfaced with a fiber channel through wireless-to-optical interfaces. Received wireless signal at each interface are sampled and quantized prior to transmission. The capacity of the FAWNA approaches the capacity of the architecture exponentially with fiber capacity. It is also shown that for a given fiber capacity, there is an optimal method of operating wireless bandwidth and number of interfaces. An optimal method to divide the fiber capacity among the interfaces is shown, which ensures that each interface is allocated a rate so that noise is dominated by front end noise rather than by quantization distortion. A method is also presented in which, rather than dynamically changing rate allocation based on channel state, a less complex, fixed rate allocation may be adopted with very small loss in performance.

The present invention relates to the compression of color image data. A combination of hard decision pixel mapping and soft decision pixel mapping is used to jointly address both quantization distortion and compression rate while maintaining low computational complexity and compatibility with standard decoders, such as, for example, the GIF / PNG decoder.

A transmit processor that includes a transmit frequency domain equalizer (TX FEQ) that pre-compensates packets in the frequency domain to flatten transmit filter response to improve the spectral mask and reduce packet error rate (PER). The TX FEQ taps may be selected so that the average power at the output equalize the average power applied to the input. The TX FEQ may be designed to yield 64 QAM packets with no quantization distortion at the input of the an Inverse Fast Fourier Transform (IFFT) processor. The 64 QAM constellation map points and TX FEQ gain values are selected to avoid exceeding the bit resolution of the IFFT processor. The tones may be bound together into zones to reduce implementation complexity by reducing the number of TX FEQ taps.

In a speaker distance calculation part of a speaker recognition apparatus, a quantization distance is obtained between a voice feature vector of a voice feature vector sequence produced from a voice of a speaker to be recognized and a representative vector of a codebook. The voice feature vector is quantized based on the quantization distance. And a quantization distortion is obtained by using a high order voice feature vector group of the voice feature vector sequence. In a recognition part of the speaker recognition apparatus, speaker recognition is executed based on the quantization distortion, which is for example, an average value of a plurality of the quantization distortions.

In a speaker distance calculation part of a speaker recognition apparatus, a quantization distance is obtained between a voice feature vector of a voice feature vector sequence produced from a voice of a speaker to be recognized and a representative vector of a codebook. The voice feature vector is quantized based on the quantization distance. And a quantization distortion is obtained by using a high order voice feature vector group of the voice feature vector sequence. In a recognition part of the speaker recognition apparatus, speaker recognition is executed based on the quantization distortion, which is for example, an average value of a plurality of the quantization distortions.

A dither system for a quantizing device, such as a multi-stage pipelined analog-to-digital converter (ADC), derives a dither signal from a clock signal having a sample frequency, the dither signal having a frequency that is one-third of the sample frequency. The dither signal is easily converted to analog and added at the input of the quantizing device to an analog signal to be digitized. A cancellation signal circuit generates a cosine-wave signal from a digital version of the dither signal and programmable coefficients that are a function of amplitude and phase. The cosine-wave signal is combined with the digital output signal from the quantizing device to produce a corrected digital output signal having reduced quantization distortion.

The invention discloses a learning-based low bit rate compressed image super-resolution reconstruction method, and the algorithm is divided into an offline part and an online part. In the offline part, the low-resolution image is first compressed with different compression quality parameter values; then the compressed image is filtered, and then the quantized distortion degree of the filtered compressed image is used as a feature, and the filtered LR image is classified according to its degree of distortion. Establish a classification sample library for multiple classes, and then use each type of sample to train the super-resolution model. In the online part, the input image is firstly filtered, and then its compression distortion category is determined. Then, according to the determination result, the sample library and super-resolution model of the corresponding category are selected to realize super-resolution restoration based on learning. Compared with other algorithms, the method of the present invention can adaptively adjust the matching sample library for input LR images with different degrees of distortion, and can effectively solve the impact of block effect distortion on image super-resolution, and directly solve the problem of low bit rate Compared with the super-resolution restoration of the distorted image, the image reconstructed by the method of the present invention has higher subjective and objective quality.

An image compression method comprises the following steps applied to at least one image: decorrelating the image by applying thereto a mathematical transform to obtain a set of coefficients, decomposing the image into blocks and, for each block of the image, quantizing the coefficients using a dead-zone uniform scalar quantizer having a semi-dead zone of size T and a quantization step size Δ, and coding the quantized coefficients, the size T of the semi-dead zone of the scalar quantizer being determined by minimizing the difference between an estimated quantization distortion D(T), dependent at least on size T, and a target quantization distortion Dc.

Disclosed are an analog-to-digital conversion method and apparatus capable of reducing quantization distortion in a receiver supporting multi-standard radios for which digital signal processing must be differently performed. The analog-to-digital conversion apparatus includes a first filter for filtering an input analog signal into a plurality of sub-band signals, a down-sampler for down-sampling the sub-band signals, a sub-band processor for allocating quantization bits for the sub-band signals, respectively, and digitizing the sub-band signals by using the allocated quantization bits, an up-sampler for up-sampling the digitized sub-band signals, and a second filter for reconstructing the sub-band signals into one digital signal. Multi-standard radio signals can be processed into digital signals in a receiver to which SDR is applied, and quantization distortion can be reduced during a procedure of converting the radio signal into the digital signal.

A FAWNA that allows high-speed mobile connectivity by leveraging the speed of optical networks. Specifically, SIMO FAWNA, which comprises a SIMO wireless channel interfaced with a fiber channel through wireless-to-optical interfaces. Received wireless signal at each interface are sampled and quantized prior to transmission. The capacity of the FAWNA approaches the capacity of the architecture exponentially with fiber capacity. It is also shown that for a given fiber capacity, there is an optimal method of operating wireless bandwidth and number of interfaces. An optimal method to divide the fiber capacity among the interfaces is shown, which ensures that each interface is allocated a rate so that noise is dominated by front end noise rather than by quantization distortion. A method is also presented in which, rather than dynamically changing rate allocation based on channel state, a less complex, fixed rate allocation may be adopted with very small loss in performance.

The invention discloses a multi-level codebook vector quantitative method for compressed encoding of a hyperspectral remote sensing image and relates to the technical field of image processing. According to the method, a spectral vector of the hyperspectral image is divided into a low-dimension part, a middle-dimension part and a high-dimension part according to the distortion condition, then a large-size codebook is adopted in the low-dimension part with large distortion, a medium-size codebook is adopted in the middle-dimension part with not large distortion, a small-size codebook is adopted in the high-dimension part with small distortion, and therefore the multi-level codebooks are adopted, the mode that only a quarter of the weight training code index of the low-dimension part is extracted after dispersion degrees are ranked is combined, and the targets of effectively reducing the quantizing distortion of the hyperspectral image and obviously reducing calculation amounts of all parts are achieved with the same compression ratio. On the condition of the low calculation complexity, high-quality compressed encoding of the hyperspectral image is achieved with higher speed, and the method has the actual application value and is a hyperspectral image lossy and nearly lossless compression scheme good in compression performance.

The invention relates to a method for processing quantization distortion information of the DCT coefficient in video coding, which comprises the steps of S1, loading a distribution model of the DCT coefficient, wherein the distribution model of the DCT coefficient is composed of a Laplacian distribution part and a uniform distribution part; S2, optimizing the uniform distribution part of the distribution model of the DCT coefficient; S3, determining a quantization distortion estimated value function DLPTCM according to the distribution model of the DCT coefficient and a quantization reconstruction method; and S4, optimizing the quantization distortion estimated value function according to the optimized tail uniform distribution part of the distribution model of the DCT coefficient. Compared with the prior art, a quantization distortion estimation algorithm is calculated for a DZ+UTQ quantization method on the basis of a quantization distortion calculation method and the distribution model of the DCT coefficient, and the quantization distortion estimation algorithm in allusion to the construction rules well solves the application thereof in other optimization algorithm fields.

A quantization device and quantization method are provided that reduce coding distortion with a small degree of calculation and achieve adequate coding performance thereby. A multistage vector quantization unit treats a number of candidates N that are designated prior to operation in the first-stage vector quantization unit, decrements the number of candidates by one beginning with the second-stage vector quantization unit and continuing with each stage thereafter. If the number of candidates is three or less, the multistage vector quantization unit assesses the quantization distortion at each stage, treating the number of candidates at the following stage as a predetermined value P if the quantization distortion is greater than a prescribed threshold, and treating the number of candidates at the following stage as a value Q that is less than the predetermined value P if the quantization distortion is less than or equal to the predetermined threshold.

The invention discloses a partition-based video steganography analysis method and system having the neighbourhood optimal probability. The method comprises the following steps of: selecting one from a partition mode set according to features of a video; for one video frame, counting a set of neighbourhood optimal SADs in all macro-block partition modes, calculating steganography analysis features of one video frame, and combining the features to obtain final steganography analysis features of one video frame; and, according to this, sequentially extracting steganography analysis features of all frames in one video. Oriented on detection of a video motion vector steganography method, aiming at the problem that the traditional prediction residual-based steganography analysis features are easily interfered by quantization distortion, the partition-based video steganography analysis method and system disclosed by the invention realize universal and precise steganography analysis to the motion vector steganography method through combination of quantization of macro-block partition to quantization distortion and neighbourhood optimal probability features.

The invention discloses an audio encoding and decoding method based on human ear auditory frequency scale signal decomposition, which is used for converting a processing result of a linear frequency domain to a corresponding human ear auditory frequency scale (Bark) to control a quantization error of a linear frequency domain parameter in order to better use a psychoacoustic model. Signal decomposition is directly carried out on the auditory frequency scale of human ears; projection calculation of a time domain signal in a Bark scale is realized; according to the invention, non-uniform sub-band decomposition of the audio signal is carried out based on the human ear auditory frequency scale ; the psychoacoustic model of human ear auditory sense is effectively utilized, and calculation precision and encoding and decoding efficiency of psychological acoustic model are improved According to the method, the quantization object is the sub-band time domain signal, the coding and decoding quantization distortion is controlled within the parameter allowable range of the psychological acoustic model on the premise that the audio coding and decoding code rate is met, the quantization mode isnovel, and the efficiency is high.