30results about How to "Guaranteed denoising effect" patented technology

A medical ultrasound image denoising method based on wavelet transform and quick bilateral filtering comprises the following steps of the first step of establishing a medical ultrasound image model, the second step of carrying out wavelet transform on an image obtained in the first step after logarithm transformation to obtain four frequency domains (LL1, LH1, HL1and HH1), continuously carrying out wavelet transform on the low frequency domain LL1 to obtain four frequency domains (LL2, LH2, HL2 and HH2) again and then repeating the step until resolving the maximum layer number, the third step of carrying out quick bilateral filtering on the low frequency portion (LLJ) at the last layer, the fourth step of carrying out threshold value method shrinkage on wavelet coefficients of the high frequency portions (LHj, HLj and HHj) of all layers and the fifth step of carrying out wavelet inverse transformation to obtain the denoised medical ultrasound image. In addition, the J is 1, 2, ..., J. If denoised ultrasound envelope signals are wanted, exponential transformation is carried out on the ultrasound image obtained in the fifth step.

The invention relates to an image de-noising method and device. According to the image de-noising method, pixels in an image are recursively filtered, and the following modes are adopted to obtain the filtering intensity of pixels to be processed currently: edge information on the chrominance component in the horizontal and vertical direction of a central pixel is calculated, a maximum value is selected as a first edge information maximum value; the variance of an image block is calculated; the maximum value of the first difference of the central pixel and filtered pixels selected from the (k-1)-th row/(k-1)-th column of the image block is calculated; edge information on the luminance component of the central pixel is calculated; whether a frequency region where the center pixel is located a flat region is determined; if the frequency region where the center pixel is located a flat region, the current brightness of the center pixel is calculated; and first filter intensity is obtained according to the current brightness and is adopted as the filtering intensity of the center pixel. With the image de-noising method and device provided by the above technical schemes of the invention adopted, the noise removal effect of noises can be ensured, and the tailing phenomenon of the de-noised image can be avoided.

The invention relates to a pulse wave motion artifact removal method for parameter adaptive optimization VMD, and belongs to the technical field of biomedical signal acquisition and processing. The method comprises the following steps of by optimizing an objective function of variational mode decomposition, evaluating the variational mode decomposition result by adopting a correlation coefficientwith the fusion of the intrinsic mode correlation degree and the central frequency dispersion degree; carrying out parameter adaptive optimization variational mode decomposition on a pulse wave signalto obtain an intrinsic mode [uk] of the pulse wave, and designing a denoising rule according to the relative energy of frequency bands corresponding to different components of the signal to remove the motion artifacts of the pulse wave. Compared with the prior art, the method has the advantages that the quality of pulse wave variational mode decomposition is improved, the effect of removing the motion artifacts is improved, and the data complexity in the process of removing the motion artifacts is simplified.

The invention discloses a rapid multi-resolution denoising method and device under a hybrid noise model. The method comprises the following steps: after acquiring an image with noise, building a multilayer gaussian pyramid and a multilayer Laplacian pyramid corresponding to theimage with the noise; performing a plurality of times of iteration calculation to obtain total variation regularization denoising target functions respectively by using a gradient descent method; performing the final iteration calculation to obtain a denoised image and the like. The device comprises a storage for storingat least one program, and a processor for loading at least one program for implementing the method. By adopting the rapid multi-resolution denoising method and device, the plurality of times of iteration calculations of the gaussian pyramid and the Laplacian pyramid are combined, the target functions can be solved corresponding to different image resolutions in each iteration calculation, the initial value of each iteration calculation is derived from the result of the previous iteration calculation and the Laplacian pyramid, denoising effect can be ensured, the calculation load of resolvingis lowered, the resolving speed is increased, and the calculation efficiency is increased.

The invention provides a frequency division diffusion filter fault strengthening method based on optimized Morlet wavelet. The method comprises the following steps: converting post-stack seismic datainto an MATLAB identifiable data format; carrying out spectral analysis on seismic data and extracting the wavelet; improving a Morlet wavelet generating function and constructing a Morlet wavelet function composition; carrying out wavelet frequency division processing by utilizing the optimized Morlet wavelet; calculating a structural tensor of the post-stack seismic data and further constructingnew fault factors based on the structural tensor; under the protection of the structural tensor, carrying out anisotropic diffusion filter on frequency division seismic data; and carrying out waveletinverse transformation on a frequency division filter result and reconstructing to obtain improved signal-to-noise ratio and resolution ratio and protect the seismic data of fault information. The method has the advantages that the resolution ratio of the seismic data is improved while denoising and edge protecting are realized, so that the low-order fault recognition ability of the seismic datais improved.

The invention relates to a method and a device for eliminating image noises. An image includes p image blocks, the size of each image block is M*N, and M and N are odd numbers. The method comprises the steps of receiving the image blocks on a Bayer domain; adopting sub-image blocks in the size of Sm*Sn with a pixel having the same channel with a central pixel as a center, wherein the central pixel is the central pixel of the received image block, and Sm and Sn are odd numbers; acquiring a filtering weight coefficient of each sub-image block, filtering the image blocks according to the filtering weight coefficient to obtain a real color difference of the central pixel as a first color difference; and calculating a pixel value of the denoised central pixel according to the first color difference and a preset denoising adjustment coefficient. By use of the above-mentioned scheme, the calculated amount of noise processing is reduced while guaranteeing denoising effects.

The invention discloses an array arrangement circuit for measuring the heart rate and blood oxygen. The circuit comprises a transmitting group, a receiving group and a sensor; the transmitting group,the receiving group and the sensor are arranged on the surface, making contact with a wrist, of a detection device; the transmitting group comprises four transmitting parts emitting different color light signals, and the transmitting parts are arranged in a rectangular array; the receiving group comprises six receiving parts receiving different color light signals, and the receiving parts are arranged in an annular array. Accordingly, by means of color light signals with different wavelengths, heart rate and blood oxygen measurement is conducted according to different measurement situations, and the whole measurement is higher in accuracy and better in stability.

The embodiment of the invention provides a voice processing method and device and a computer storage medium, and the method comprises the steps: carrying out the duration division of inputted noisy voice information and annotation data, and obtaining at least one piece of sub noisy voice information and at least one piece of sub annotation data; training a discrimination model and a generation model based on the at least one piece of sub-noisy voice information and the at least one piece of sub-annotation data to obtain a trained discrimination model and a trained generation model, wherein thegeneration model is used for carrying out denoising processing, and is a network comprising at least N convolution layers.

The invention discloses a magnetic resonance image denoising method which comprises the following steps: performing tensor decomposition; splitting image blocks; calculating the similarity weight; estimating the pixel value; and performing deviation correction. The magnetic resonance image denoising method disclosed by the invention has the advantages that the magnetic resonance image denoising effect is improved, so that more original image information can be saved, and blurring of the image edge part is reduced. The magnetic resonance image processing speed is improved. The method can be combined with other improved non-local mean denoising algorithms and is used for improving the image processing speed, and the denoising result can be ensured.

The invention discloses a quick X-ray dynamic real-time imaging denoising method based on deep learning. The method comprises the following steps: S1, preprocessing clinically collected X-ray dynamic real-time video data to obtain a data set; s2, a deep learning model comprising at least one denoising module is constructed, and the at least one denoising module forms a layer of network or a cascaded two-layer network; s3, training the deep learning model by using the data set; and S4, inputting a plurality of continuous frames of the X-ray dynamic real-time video with low radiation dose into the deep learning model trained in the step S3, and outputting a denoising result of a middle frame of the plurality of continuous frames. According to the deep learning-based fast X-ray dynamic real-time imaging video denoising method provided by the invention, the denoising time is greatly shortened on the basis of ensuring the denoising effect, and the deep learning-based fast X-ray dynamic real-time imaging video denoising method can be applied to operation video real-time denoising processing which cannot be achieved by the conventional method.

The invention aims to provide a wavelet threshold denoising parameter selection method based on a composite evaluation index and wavelet entropy, which comprises the following steps: acquiring a standby wavelet basis, carrying out wavelet denoising processing, calculating the composite evaluation index, respectively carrying out normalization processing on a data set consisting of two traditional evaluation indexes of different decomposition layers of each wavelet basis, and selecting a wavelet threshold denoising parameter according to the normalization processing. Each wavelet basis obtains a group of composite evaluation indexes, the optimal decomposition layer number is determined, the similar wavelet basis is used for carrying out layer-by-layer wavelet decomposition on a noisy signal to the optimal decomposition layer number, the optimal wavelet basis of each layer of decomposition is determined by calculating the wavelet entropy of a low-frequency coefficient, and the composite evaluation indexes are compared to determine the optimal denoising scheme in the multiple types of wavelet basis. According to the method, the optimal decomposition layer number is determined by constructing the composite evaluation index, and the optimal wavelet basis of each layer of decomposition is determined by calculating the low-frequency coefficient wavelet entropy, so that the optimization problem of two denoising parameters, namely the wavelet basis and the decomposition layer number, in wavelet denoising is solved on the premise that the calculation amount is reduced as much as possible.

The invention belongs to the field of digital image processing, and particularly relates to a surface point cloud reconstruction method based on binocular stereoscopic vision. The method includes the following steps: step 1, performing stereo correction on images shot by a binocular camera to enable left and right images to be in the same pole with roll call; 2, carrying out the preprocessing of the corrected image; 3, carrying out complex background removal on the region of interest through a minimum cut-maximum flow image segmentation algorithm; 4, recovering depth information through a convolutional neural network stereo matching algorithm to obtain a disparity map; and 5, reconstructing the surface point cloud according to the disparity map obtained in the step 4. According to the method, the problems of low reconstruction precision, slow speed, poor transferability and the like are solved through processes of stereo correction, image preprocessing, region-of-interest background removal, stereo matching, point cloud reconstruction and the like.

The invention discloses an MEMS gyroscope signal denoising method based on a self-adaptive multi-scale filter. The method comprises the following steps: acquiring MEMS gyroscope signals, and calculating sample variances of the signals at sampling points; adaptively adjusting the left window length and the right window length of each data point in the SG filter according to the sample variance of each data point in the signal and preset maximum and minimum window lengths; and performing adaptive multi-scale SG filtering on each sampling point signal by using the obtained left and right window lengths of each sampling point. In order to realize rapid compensation for random errors of MEMS gyroscope signals, a mapping model is established for sample variance based on adaptive moving average and the window length of an SG filter, and the window length of the SG filter is adaptively adjusted, so that de-noising of non-stationary and non-linear gyroscope signals is realized, and the multi-scale window length enables the filter not to cause signal deformation while achieving a good denoising effect.

The invention discloses a heart rate analysis method and system based on a heart sound signal, and the method comprises the steps: carrying out initial filtering processing of an original heart sound signal through a band-pass filter, and carrying out the time delay of the processed original heart sound signal, and obtaining a heart sound reference signal; defining an adaptive filter and determining a cost function used by the adaptive filter; inputting the processed original heart sound signal and the processed heart sound reference signal into the adaptive filter, and updating a weight vector of the adaptive filter by taking the minimum cost function as a criterion; calculating a weight change vector of the adaptive filter at the current moment, and determining an effective boundary range by using a comparison result between the weight change vector and a set threshold value; screening out a plurality of heart sound segments from the effective boundary range based on a set minimum segment length, and further constructing an effective heart sound signal; conducting envelope extraction on the effective heart sound signal based on a single-degree-of-freedom model, and then calculating heart rate parameters corresponding to the effective heart sound signal. The embodiment of the invention can improve the accuracy of a heart rate measurement result.