530 results about "Upsampling" patented technology

Adaptive up-sample filtering is used to improve compression efficiency of spatially scalable coding systems by more effectively predicting the high-resolution (enhanced-layer) video (or image) from the low-resolution lower-layer video (or image). Different up-sample filters adaptive to local image properties are selectively used for different portions of a low resolution frame to generate a better up-sampled image. Selection between different up-sample filters is determined by a variety of different information available to both the encoder and decoder. In one embodiment, the up-sample filters are selected by the encoder and then explicitly identified to the decoder. Other techniques are then used to minimize the cost of transmitting the up-sample filter identifiers. In alternative embodiments, the encoder and decoder independently make up-sample filters selections.

Video encoding and decoding methods and systems for video streaming are provided. The video encoding method includes encoding first resolution frames using scalable video coding, upsampling the first resolution frames to a second resolution, and encoding second resolution frames using scalable video coding with reference to upsampled versions of the first resolution frames.

There are provided scalable complexity video decoder and encoders for respectively decoding and encoding a video bitstream. A scalable complexity video decoder includes a first combiner, a second combiner, and a deblocking filter. The first combiner is for combining full resolution motion compensated predictions and decoded upsampled base layer residuals to form reconstructed upsampled base layer pictures for display without any deblocking operations applied thereto. The second combiner, in signal communication with the first combiner, is for combining the reconstructed upsampled base layer pictures with decoded enhancement layer residuals to form reconstructed full resolution enhancement layer pictures. The deblocking filter, in signal communication with the second combiner, is for performing deblocking operations on only the reconstructed full resolution enhancement layer pictures to output filtered reconstructed full resolution enhancement layer pictures for display.

A scalable video bitstream may have an H.264/AVC compatible base layer and a scalable enhancement layer, where scalability refers to color bit depth. The H.264/AVC scalability extension SVC provides also other types of scalability, e.g. spatial scalability where the number of pixels in BL and EL are different. According to the invention, BL information is upsampled in two logical steps, one being texture upsampling and the other being bit depth upsampling. Texture upsampling is a process that increases the number of pixels, and bit depth upsampling is a process that increases the number of values that each pixel can have, corresponding to the pixels color intensity. The upsampled BL data are used to predict the collocated EL. The BL information is upsampled at the encoder side and in the same manner at the decoder side, wherein the upsampling refers to spatial and bit depth characteristics.

Systems and methods for automated segmentation of anatomical structures, such as the human heart. The systems and methods employ convolutional neural networks (CNNs) to autonomously segment various parts of an anatomical structure represented by image data, such as 3D MRI data. The convolutional neural network utilizes two paths, a contracting path which includes convolution/pooling layers, and an expanding path which includes upsampling/convolution layers. The loss function used to validate the CNN model may specifically account for missing data, which allows for use of a larger training set. The CNN model may utilize multi-dimensional kernels (e.g., 2D, 3D, 4D, 6D), and may include various channels which encode spatial data, time data, flow data, etc. The systems and methods of the present disclosure also utilize CNNs to provide automated detection and display of landmarks in images of anatomical structures.

A system and method for semantic segmentation using dense upsampling convolution (DUC) are disclosed. A particular embodiment includes: receiving an input image; producing a feature map from the input image; performing a convolution operation on the feature map and reshape the feature map to produce a label map; dividing the label map into equal subparts, which have the same height and width as the feature map; stacking the subparts of the label map to produce a whole label map; and applying a convolution operation directly between the feature map and the whole label map without inserting extra values in deconvolutional layers to produce a semantic label map.

A method of more efficiently conducting temporal filtering in a scalable video codec by use of a base-layer is provided. The method of efficiently compressing frames at higher layers by use of a base-layer in a multilayer-based video coding method includes (a) generating a base-layer frame from an input original video sequence, having the same temporal position as a first higher layer frame, (b) upsampling the base-layer frame to have the resolution of a higher layer frame, and (c) removing redundancy of the first higher layer frame on a block basis by referencing a second higher layer frame having a different temporal position from the first higher layer frame and the upsampled base-layer frame.

A method upsamples an image using a non-linear fully connected neural network to produce only global details of an upsampled image and interpolates the image to produce a smooth upsampled image. The method concatenates the global details and the smooth upsampled image into a tensor and applies a sequence of nonlinear convolutions to the tensor using a convolutional neural network to produce the upsampled image.

A novel and useful apparatus for and method of upsampling/interpolating a discrete-time input sample stream with time alignment utilizing the addition of randomized high frequency noise. The upsampling mechanism is an effective implementation of a second order interpolator that eliminates the need for a conventional filter as the filtering action is effectively built into the mechanism. The upsampling mechanism takes the derivative of the discrete-time input sample stream, thereby effectively providing another order of interpolation over a conventional interpolator. Before outputting the interpolated signal, an integrator takes the integral of the interpolated samples. Any processing performed between the derivative and integrator blocks effectively provides an additional order of interpolation. High frequency noise (i.e. dithering) is added to the differentiated samples in order to eliminate the spectral regrowth spurs that would otherwise appear in the output after rounding. Delay alignment is performed on the differentiated samples in order to time align both phase/frequency and amplitude samples that are processed on different paths.

The invention discloses a single-image super-resolution reconstruction method based on a deep residual network. The method of the invention mainly comprises a first step of performing block extraction and pixel averaging processing on an image in a sample image database to obtain a corresponding high resolution and low resolution training image sets; a second step of constructing a deep convolutional neutral network with a residual structure for iterative training, and then inputting the training set obtained in the first step to the neural network constructed in the second step for iterative training; and a third step of according to a data model obtained by training, realizing the continuous up-scaling of the input low resolution image through the combination of iterative operation and an interpolation algorithm. By introducing a deep residual network and introducing an upsampling layer at the end of the network, the method of the invention accelerates the processing speed of the image up-scaling, enhances the display effect of the image details, obtains a better image super-resolution reconstruction effect, and has a wide range of applications in the image high definition display, image compression, security checks and other fields.

Systems and methods for upsampling input images may evaluate potential upsampling solutions with respect to an objective function that is dependent on a sparse derivative prior on second derivative(s) of the potential upsampling solutions to identify an acceptable higher-resolution output image. The objective function may also be dependent on fidelity term(s) and/or sparse derivative prior(s) on first derivative(s) of potential upsampling solutions. The methods may include applying the iteratively re-weighted least squares procedure in minimizing the objective function and generating improved candidate solutions from an initial solution. The identified solution may be stored as a higher-resolution version of the input image in memory, and made available to subsequent operations in an image editing application or other graphics application. The methods may produce sharp results that are also smooth along edges. The methods may be implemented as program instructions stored on computer-readable storage media, executable by a CPU and/or GPU.

A system and method is disclosed for reconstructing high frequency components of a digital audio signal using a harmonic enhancer in a baseband integrated circuit of a receiver handset. The original spectrum of the digital audio signal is upsampled in a times two (2) upsample unit to double the size of the bandwidth. A low pass filter then removes a high frequency alias of the original spectrum. The spectrum is then modulated with a first carrier frequency and sent to a first filter bank where a low pass filter and a high pass filter shape the modulated harmonic spectrum. After gain adjustment, the modulated harmonic spectrum is added to a delayed version of the original spectrum. Additional harmonic spectra are similarly created at other carrier frequencies and added to the audio output spectra to reconstruct high frequency components of the audio signal.

A system method for occluding contour detection using a fully convolutional neural network is disclosed. A particular embodiment includes: receiving an input image; producing a feature map from the input image by semantic segmentation; applying a Dense Upsampling Convolution (DUC) operation on the feature map to produce contour information of objects and object instances detected in the input image; and applying the contour information onto the input image.

The invention discloses a remote sensing image semantic segmentation technology based on a VGG network. The method comprises the following steps: 1, randomly cutting the high-resolution remote sensingimage for training and the corresponding label image into small images; dividing the network structure into an encoding part and a decoding part; adopting the depooling path and the deconvolution path to double the resolution of the coded information; carrying out channel connection on a characteristic image and a result of cavity convolution, recovering the characteristic image to an original size through deconvolution upsampling, inputting an output label image into a PPB module for multi-scale aggregation processing, and finally, updating network parameters in a random gradient descent mode by taking cross entropy as a loss function; inputting the small images formed by sequentially cutting the test images into a neural network to predict corresponding label images, and splicing the label images into an original size. According to the technical scheme, the segmentation precision of the model is improved, the complexity of the network is reduced, and the training time is saved.

The invention discloses a single image super-resolution reconstruction method combining depth learning and gradient transformation. The method comprises the steps that a super-resolution method based on depth learning is used to carry out upsampling on an input low-resolution image to acquire an upsampling image; a gradient operator is used to carry out gradient-extracting on the upsampling image; a depth convolutional neural network is used to convert extracted gradient; a cost function is reconstructed by using the input low-resolution image and the converted gradient as constraints; a gradient descent method is used to optimize the reconstructed cost function to acquire a final output high-resolution image. According to the single image super-resolution reconstruction method provided by the invention, the reconstructed image has a fine structure in the subjective visual effect, is free of artificial effect, and has a high objective evaluation parameter value. The invention provides the effective single image super-resolution reconstruction method.

The invention relates to a virtual viewpoint image generation method based on a generation type confrontation network. The method comprises the following steps: first step, making a data set: obtaining an image pair necessary for training the generation type confrontation network; constructing a model: the structures used by a generator and a discriminator are that batch normalization layers BatchNorm and nonlinear operation unit ReLU activation functions are connected behind convolutional layers, all convolutional layers use a 4*4 convolution kernel size, the step size is set as 2, the lengthand width are both reduced to a half of the original length and width when downsampling is performed on a feature image, the length and width are both amplified to twice of the original length and width during upsampling, and a Dropout layer sets a Dropout rate as 50%; the RelU activation function is LeakyReLu; defining the loss; and training and testing the model.

The invention discloses an instance segmentation method based on feature attention and sub-upsampling. The method comprises the steps of obtaining a to-be-segmented original image; extracting a feature map from the original image through a feature global network, and determining a region of interest in the feature map, the feature global network comprising an attention module; aligning and extracting a region of interest from the feature map; and classifying the extracted regions of interest, and generating segmentation masks for the extracted regions of interest by utilizing sub-pixel up-sampling so as to realize instance segmentation of the original image. The method has the advantages that the attention module is added when the feature map is extracted; according to the embodiment of the invention, redundant information and fusion information can be deleted by applying channel transformation after maximum pooling and average pooling operations, the effectiveness of an image featurevector is improved. Meanwhile, the accuracy of segmentation and detection in instance segmentation is improved under the condition of not reducing the speed in combination with a sub-pixel up-samplingmode, and the occupied memory is not increased.

The invention relates to a rectangular-target identification method based on multiple channels and multiple threshold values. The identification method comprises the following steps of removing image noise from an original image through pyramid downsampling at first and then pyramid upsampling; extracting each channel image in the original image; setting N gray levels, and conducting binaryzation on each channel image by making the threshold values Tg be equal to 255/n; conducting region labeling on the binarized images, and extracting the outline of each region; using polygons for approximating each peripheral outline, and recording the coordinates of vertexes of each polygon; selecting the polygon with the number of vertexes being four, and calculating each interior angle of the quadrangle; deterring that the region surrounded by the quadrangle is a rectangular target if each interior angle of the quadrangle approximates 90 degrees. The rectangular-target identification method can solve the detection problem that rectangular targets in various colors exist in one image; meanwhile, the multiple threshold values are adopted for conducting image binaryzation, and the detection accuracy rate of the rectangular targets is greatly increased.

An all-digital class-D audio amplifier and method of digitizing an analog signal comprises a pulse-width modulation (PWM) wave generator for generating a PWM wave; an interpolator for receiving an input audio signal, up-sample the input audio signal; and down-sample the input audio signal to a PWM wave switching frequency; an audio speaker for broadcasting an amplified version of the input audio signal; a pair of MOSFETs driven by the PWM wave and the input supply voltage, wherein the pair of MOSFETs transmit the amplified version of the input audio signal to the audio speaker; a natural-sampling and pre-distortion circuit adapted to (i) generate a pre-distorted natural-sampling frequency point for the PWM wave switching frequency, and (ii) account for digitized power supply ripples for an input supply voltage used to drive the audio speaker; and an analog-to-digital converter (ADC) adapted to digitize the power supply ripples or distortion.