32 results about "Wavelet coding" patented technology

A new effective and fast method and apparatus for still image compression implements an embedded progressive sorting scheme in a quadtree-like structure. In contrast to zerotree-based methods for wavelet coding, the invented embedded quadtree wavelet (EQW) method exploits the inherent spatial self-similarity within individual layers of the multiresolution decomposition hierarchy. This self-similarity offers higher predictability of the data within the same resolution level, and therefore usually provides a higher performance in seeking a compact code. The computation involved in the EQW method is more efficient than in the zerotree wavelet coding, and the produced bitstream is more robust to channel noise. The present invention can effectively be used for object-oriented shape coding or region coding in image and video compression coding systems.

Provided is an inter-frame wavelet coding apparatus that can reduce the computation complexity of a decoder by adjusting a decomposition level and a filter length based on the information amount of a frame during wavelet transform and a method therefor. The inter-frame wavelet coding apparatus includes: a Motion Compensated Temporal Filtering (MCTF) unit for computing a motion vectors of a group of pictures (GOP) and filtering the GOP with respect to the temporal axis, to thereby obtain filtered frame; a wavelet transforming unit for performing spatial wavelet transform on the filtered frame and outputting a wavelet coefficient; a quantization unit for quantizing the wavelet coefficient; an entropy coding unit for entropy-coding the motion vector computed in the MCTF unit and the quantized wavelet coefficient, to thereby generate an entropy-coded bit stream; and a wavelet filter managing unit for selecting a decomposition level and a filter length for the wavelet transforming unit based on motion estimation information of the GOP video computed in the MCTF unit, wherein the decomposition level and the filter length are included in the entropy-coded bit stream.

The invention relates to a three-dimensional point cloud model data compression method based on four-side patch segmentation. The method comprises the following steps of (1), segmenting the surface ofan input point cloud model to a plurality of quadrangular patches; (2), successively performing parameterization and resampling processing on the quadrangular patches, and obtaining two-dimensional image data; performing regular resampling on the patches, wherein after resampling, each patch can be described by a height picture, thereby converting three-dimensional data to two-dimensional image data and reducing data volume; (3), performing wavelet coding on the two-dimensional image data by means of an EZW algorithm, and realizing progressive compression of a three-dimensional point cloud model; and (4), performing decoding and model reconstruction. An existing OT algorithm has a remarkable sawtooth phenomenon on the condition of low bit rate, and the method according to the invention well settles the defect. On the condition of the same bit rate, the method according to the invention realizes a better visual effect. The method of the invention increases three-dimensional model datacompression rate.

The invention discloses an asymmetrical wavelet-base based hyper-spectral image wavelet compression method and a device thereof. The device consists of a wavelet base selection and evaluation module, an asymmetrical wavelet transformation module, a wavelet coding module, a compression data transmission and storage module, a wavelet decoding module and an asymmetrical wavelet reverse transformation module. Based on a decoupling wavelet transformation structure, the method respectively adopts different wavelet bases for transforming spectral dimension and spatial dimension data and adopts a dimension-reduced wavelet base selection and evaluation mechanism to confirm the optimal wavelet base for the spectral dimension and spatial dimension of hyper-spectral image. The method and the device provided by the invention can strengthen the coefficient characteristics after wavelet transformation and then improve the compression performance of wavelet coding. Compared with the traditional wavelet transformation, the wavelet compression based on asymmetrical wavelet base and asymmetrical structure can not add calculation complexity or affect the expansibility after compression.

The invention provides a strip wave transform image compression method based on a graphics processor. The method overcomes the defects that in the prior art, the calculation quantity is large and the peak signal to noise ratio of compressed images is low. The method specifically includes the following steps of firstly, inputting an image to be compressed; secondly, conducting wavelet transform; thirdly, conducting Bandelet conversion; fourthly, conducting embedded zerotree wavelet coding and outputting a compressed file; fifthly, inputting the compressed file to be decompressed; sixthly, conducting decoding; seventhly, conducting Bandelet inverse transformation and outputting a reconstructed image. The method can effectively solve the problems that in the prior art, the compression speed is low, image detail characteristics can not be well kept and the peak signal to noise ratio is low, the detail characteristics of the image can be well reserved, the compression speed is increased, and the peak signal to noise ratio of the compressed image is increased.

Compression of images that have masked or “don't care” regions which are delineated by a binary image mask is achieved using “masked wavelet transforms.” A unique mask-dependent lifting scheme is used to compute invertible wavelet transforms of the input image for use in encoding and decoding the input image. These mask-dependent wavelet transforms are derived from the input image based on the masked regions within the image. Masked wavelet coding automatically generates an appropriate linear combination of available, unmasked, neighboring pixels, for both the prediction and the update steps of “lifting” for each pixel. This pixel availability is then used to change the wavelet function on a case-by-case basis as a function of the mask by using a polynomial of degree k−1 for interpolation in both the predict and update steps of lifting where at least k unmasked neighboring pixel values are available.

The invention discloses a fractal-wavelet self-adaptive image denoising method based on a multivariate statistic model. The method includes: step one, subjecting a noisy image to homomorphic transform through which an original image IB containing multiplicative noise is transformed into an image IB' only containing additive noise; step two, performing fractal-wavelet transform on a noisy signal f (k), selecting a wavelet basis and a wavelet decomposition layer j to obtain corresponding wavelet coefficients; step three, selecting an MGGD multivariate statistic model for self-adaptive solution of a parameter alpha and a parameter beta, and obtaining the most suitable parameter value alpha and beta after analysis for the distribution condition of the wavelet coefficient of a natural image; step four, for the wavelet coefficients obtained through decomposition, performing noise-free predictive coding on the noisy image by using a fractal-wavelet coding method; and step five, performing wavelet reconstruction by using the wavelet coefficients to obtain estimation signals which are image signals after denoising. Compared with other algorithms, the method has better denoising effect and high edge preserving capacity, and is particularly suitable for eliminating Gaussian-impulse mixed noise.

Wavelet coding of partially-masked image information may be made faster by either of two embodiments of the present invention. In a first embodiment, quick convergence is obtained by performing wavelet encoding in stages, each stage associated with a predetermined wavelet scale. By advancing the stages from finest scale to coarsest scale, coefficients of masked wavelets are identifies early in the coding process. In a second embodiment, quick convergence is obtained by introducing overshoot techniques when masked coefficients are identified, modified and image data is reconstructed therefrom.

The hierarchical coding process has an image which is associated with a mesh, and which is coded by wavelets. Two types of wavelets are applied selectively to distinct zones of the image.

The invention provides a medicinal ultrasonic image compressing method based on wavelet coding, the method comprises: 1) peforming wavelet coding on a medicinal ultrasonic image I, peforming wavelet transformation on an imaging region, i.e., WIA (I*MI), and peforming wavelet transformation on a background region, i.e., WBG(I*(1-MI)/S), wherein MI represents a shape mark of the imaging region, function W represents integer lifting wavelet transformation, the lower corner sign IA represents a wavelet which performs transformation on the imaging region, the lower corner sign BG of the function W represents a wavelet which performs transformation on the imaging region, S is a size parameter; 2) calculating a wavelet domain mark MW based on the shape mark MI of the imaging region of the ultrasonic image, calculating a mixed wavelet parameter by the wavelet domain mark MW; 3) encoding the mixed wavelet parameter by a wavelet parameter encoding algorithm and based on a gather dividing algorithm of a delamination tree, obtaining an image compressing result. The invention improves compressing rate of the image while ensuring that information in a medicinal ultrasonic image diagnosis region has no damage, and has a wide application.

The invention discloses an EZW data compression method based on random threshold adjustment. The method comprises the following steps of: carrying out wavelet coding on an image to obtain wavelet transform; quantification, coding, decoding process is opposite thereto, when image scanning coding is carried out, an initial threshold value is calculated firstly, then EZW coding is carried out, the EZW coding adopts a method of randomly adjusting the threshold value for judging the importance of a wavelet coefficient to code a digital image, and the specific method is that when the threshold valueis divided by 2 every time, a positive random number in an adjustable range is added to a divisor 2. According to the method, the threshold value for judging the importance of the wavelet coefficientis randomly adjusted. According to the technical scheme, the threshold value is randomly selected in a small range, the compression ratio is increased while the restoration quality is ensured, the threshold value is randomly selected in different ranges for encoding the same image, the compression ratio is increased, the restoration effect of the distorted compressed image is good, and although the peak signal-to-noise ratio is reduced to a certain extent, the influence is not large.

The invention discloses embedded zerotree wavelet coding method and device. The embedded zerotree wavelet coding method comprises the following steps of: decomposing an input image into a plurality of wavelet decomposed subbands; adding all wavelet coefficients of the plurality of wavelet decomposed subbands to a primary wavelet coefficient table, and setting a secondary wavelet coefficient table to be empty; coding all the wavelet coefficients of the plurality of wavelet decomposed subbands bit plane by bit plane in two stages of a primary control procedure and a secondary control procedure according to a sequence from a highest bit plane to a lowest bit plane, wherein processing all the wavelet coefficients of the plurality of wavelet decomposed subbands one by one in the bit plane n; in the stage of the primary control procedure, judging whether the absolute value of one wavelet coefficient of the primary wavelet coefficient table is larger than or equal to 2n or not, and if so, generating a POS or NEG sign according to a sign bit of the wavelet coefficient; adding the wavelet coefficient to the secondary wavelet coefficient table; and setting the wavelet coefficient as zero in the primary wavelet coefficient table.

The present invention relates to a method for encoding a video signal and apparatus therefor, and more particularly, to a method and an apparatus therefor, the method for determining, when the number of significant bits of wavelet coefficient information of a current precinct is greater than a truncation position, a bit budget for code information of the wavelet coefficient information as 1, determining, when the number of significant bits of the wavelet coefficient information is smaller than the truncation position, the bit budget of the code information of the wavelet coefficient information as 0, performing, when the number of significant bits is same as the truncation position, quantization on the wavelet coefficient information, determining, when a result of the quantization is 0, the bit budget for the code information of the wavelet coefficient information as 0, and determining, when the result of the quantization is not 0, the bit budget for the code information of the wavelet coefficient information as 1.

The invention relates to a method which is used for converting a JPEG image into a wavelet compression image. The method comprises the steps that entropy decoding and inverse quantization are carried out on the input JPEG image to acquire a DCT coefficient matrix; the DCT coefficient matrix is split into a number of 8*8 matrixes, so as to acquire a number of DCT coefficient blocks; direct current and alternating current coefficients are extracted from each DCT coefficient block; all direct current coefficients are used as a low frequency subband, and all alternating current coefficients are used as a high frequency subband; a wavelet coefficient matrix is formed according to the low frequency subband and the high frequency subband; and wavelet coding is carried out based on the wavelet coefficient matrix, so as to acquire the wavelet compression image. The invention further provides a corresponding JPEG image and wavelet compression image conversion system, and corresponding method and system for converting the wavelet compression image into the JPEG image. According to the invention, the efficiency of image transcoding can be significantly improved; the conversion time is reduced; and the methods and the systems are suitable for fast conversion between mass JPEG images and wavelet compression images.

The invention relates to the technical field of medical image classification, and particularly discloses a medical image classification method combining wavelet transform and a tensor network. A coarse graining network is provided, the coarse graining network combines wavelet transform and MERA, that is, D4 wavelets are coded into the MERA to form a Wavelet MERA model with fixed internal parameters; and a tensor classification network like a full connection layer is also constructed. According to the invention, the MNIST data set, the Covid-19 data set and the LIDC data set are used for multi-dimensional verification, and the result shows that the accuracy stability of the wavelet MERA is high, and the wavelet MERA has better coarse graining ability than a CNNs deep neural network, so that the wavelet MERA can reduce the parameter quantity of the model to a greater extent while the precision is ensured. The result shows that the Wavelet MERA is superior to the current mainstream deep neural network in classification and is also superior to the common wavelet transform in the aspect of data preprocessing. In addition, the wavelet MERA also has the advantage of interpretability of the tensor network itself.

The invention discloses a method for controlling rate of wavelet-encode, including: obtaining the predicted rate of the current image frame which corresponds to the preset quantization step-length. If the predicted rate can not meet the preset condition which consists of bit-value, average bit-value of frame, threshold of frame-skipping and / or rate of encode frame occupied with the buffer zone, the quantitative step-length mentioned above is adjusted to meet the precondition and the preset rate meeting the condition is taken as the target one when the quantitative step-length meets the preset range. The three components of non-zero of the current image frame multiply with the multiplier factor of the corresponding image frame, add the products together, add the number of bits used for encoding of motion vector and encoding stream organizations, the result of the above is taken as the rate forecast of the current image frame.

The invention provides an image conversion method and device. The method comprises the steps that entropy decoding and inverse quantization processing are performed on compressed code streams of JPEG images to obtain DCT coefficient matrices; the DCT coefficient matrices are restructured to obtain wavelet packet coefficient matrices; inverted wavelet packet conversion processing is performed on the wavelet packet coefficient matrices to obtain wavelet coefficient matrices; wavelet coding is performed on the wavelet coefficient matrices to obtain wavelet compressed images. The image conversion method and device are capable of enhancing the speed and the efficiency of image conversion remarkably and shortening the time consumed in the image conversion process greatly, thereby being suitable for fast conversion between the large-batch JPEG images and the wavelet compressed images.

The invention relates to the field of image processing, in particular to an image compression method for medical image storage, which comprises the following steps of: acquiring a medical image, extracting a region of interest, obtaining adjacency according to the gray difference and variance of each edge pixel point of the region of interest and eight neighborhood pixel points of the region of interest; edge pixel points needing to be expanded are determined according to an adjacency threshold value, external expansion is carried out through continuous iteration, an expanded area is obtained, the combined variance of the expanded area and the combined variance of the residual area are obtained through the gray values of the pixel points of the interested area, the pixel points of the expanded area and the pixel points of the residual area, and a random area in the residual area is determined; wavelet coefficients of random regions in the region of interest, the expanded region and the remaining regions and other regions in the remaining regions are obtained through the mask algorithm, embedded zero-tree wavelet coding is carried out in sequence, image compression is completed, and the method is intelligent and efficient.