51results about How to "Good denoising" patented technology

The invention discloses a backlight image enhancement and denoising method based on foreground-background separation. The backlight image enhancement and denoising method comprises the steps that a backlight image is divided into a foreground area and a background area by adopting an interactive cutout algorithm; the pixel points in the foreground area are enhanced by adopting an improved Retinex algorithm; equalization processing is performed on the pixel points in the background area by adopting a CLAHE algorithm; denoising is performed on the foreground area after enhancement and the background area after equalization processing by adopting a multi-scale NLM algorithm; and weighted fusion is performed on the foreground area and the background area after denoising so that an enhanced and denoised backlight image is obtained. Different enhancement and denoising methods are respectively adopted for the foreground area and the background area of the backlight image so that detail enhancement of the foreground area of the backlight image can be realized, the background area can be protected from being excessively enhanced, the denoising effect is great and accuracy is high and thus the backlight image enhancement and denoising method can be widely applied to the field of backlight image processing.

The invention relates to a saliency detection and LSD linear detection-based airport extraction method comprising the following steps: an original airport detection data set is built; remote sensing images are extracted one by one from the data set and are subjected to downsampling operation, Frequency-tuned method based saliency detection is conducted on the downsampled images, a full resolution saliency map is formed, linear segments in the saliency map are obtained via LSD linear segment detection operators, the obtained linear segments are segmented and then connected to form long linear segments, parallel straight lines are screened out, obtained parallel straight lines are clustered, a minimum bounding rectangle zone for each type is extracted and then expanded outward by 1.1 times to form a final airport zone, and coordinates of the airport zone are output and drawn up in original remote sensing images. Via LSD linear detection in the airport extraction method disclosed in the invention, texture and edge information in the images can be successfully extracted, FT saliency detection is combined with the LSD linear detection, detection is expedited, and precision is ensured while detection accuracy is enhanced.

The invention discloses a multi-beam point cloud data denoising method considering terrain characteristics. A basic thought of the method is that a topological relation among point cloud data is established based on a KD index tree; near neighbor data of points fits local planes based on an RANSAC algorithm; the distances between point cloud and the local fitting planes are calculated; and denoising is performed based on a statistic analysis method. In addition, a pre-judgment is performed according to normal vector characteristics of adjacent planes before denoising to remove an obvious outlier surface, and the point cloud at an abrupt slope is reserved, thereby preventing excessive denoising. Through the method, near-surface noise and obvious outlier noise data in the multi-beam point cloud data can be removed, and information of edges and the like is better reserved; and the design scheme is optimized on the basis of ensuring the abovementioned effects, so that the executive efficiency is improved.

The invention provides a multifunctional image processing method based on wavelet transform. The multifunctional image processing method comprises the following steps: step 1, reading an original image; 2, decomposing the original image into a high-frequency part and a low-frequency part by wavelet transform; 3, for the high-frequency part of the image, performing threshold quantization processingon all high-frequency coefficients, and then performing median filtering to complete compression of the high-frequency part and image enhancement; 4, for the low-frequency part of the image, enhancing a low-frequency coefficient by adopting an improved function; and step 5, reconstructing the processed high-frequency part and the processed low-frequency part by using wavelet inverse transformation to obtain a reconstructed image. According to the method, wavelet transformation is adopted to process the image, so that the entropy after signal transformation is reduced, the non-stationarity ofthe signal can be well described, and feature extraction and protection are facilitated. According to the method, wavelet transform is adopted, so that denoising is more facilitated in a wavelet domain than in a time domain, and different wavelet functions can be selected according to different application requirements to obtain an optimal processing effect.

The invention relates to a real-time on-line system-based field leaf image edge extraction method and a real-time on-line system-based field leaf image edge extraction system. The method comprises the following steps of: 1, performing gray-level mapping processing on segmentation area sub-images; 2, performing smoothening and noise removing, thresholding, edge detection and preliminary optimization on the images processed by the step 1, and judging whether a target area has a leaf shape; 3, performing smoothening and noise removing, thresholding negation, edge detection and preliminary optimization on the images processed by the step 1, and judging whether the target area has a leaf shape; 4, performing smoothening and noise removing, thresholding, thresholding negation, edge detection, preliminary optimization and area combination on the images processed by the step 1, and judging whether the target area has a leaf shape; and 5, performing further optimization and closed operation onthe images processed by the steps 2, 3 and 4 to obtain edge-closed edge images. Through the method and the system, complete, enclosed and accurately positioned target boundary curve images can be acquired, and the segmentation success rate is high.

The invention discloses a two-dimension analysis thinning model and a dictionary training method and an image denoising method based on the two-dimension analysis thinning model, wherein the two-dimension analysis thinning model fully utilizes the spatial correlation of images, needs fewer training samples, and greatly saves the storage space of a dictionary.

The invention discloses a method for image processing based on a matrix variable variation self-encoder so that a problem of damaging the spatial structure by image vectorization processing can be solved and thus image reconstruction, denoising and completion can be realized. Different from the traditional VAE, the method has the following advantages: the input, hidden layer features, latent variable distribution parameters of the model are described by using an inherent expression form -2D matrix of the image; with the definition of matrix Gaussian distribution and related properties, an explicit expression of an objective function of a novel model is deduced and then model parameters are calculated by using a stochastic gradient descent algorithm. In the model, modeling of the spatial structure and statistical information of the image data is carried out well because the involved modeling process is oriented at the matrix variable, so that the image reconstruction quality is enhanced, noises are removed, and the image completion is realized.

The invention relates to a space-time domain joint noise estimation system. The system comprises a motion estimation module, a scale motion detection module, a global motion detection module, a scene change detection module, a time domain noise estimation module, a space domain noise estimation module and a fusion module, wherein the motion estimation module calculates and outputs a motion vector according a previous frame and a current frame; the scale motion detection module calculates and outputs scale reliability according to a motion vector; the global motion detection module calculates and outputs global reliability according to the motion vector; the scene change detection module calculates and outputs scene change reliability according to the motion vector; the time domain noise estimation module calculates and outputs a time domain noise level according to the previous frame and the current frame; the space domain noise estimation module calculates and outputs a space domain noise level according to the current frame; the fusion module calculates and outputs a final noise level according to the scale reliability, the global reliability, the scene change reliability, the space domain noise level and the time domain noise level. The system provided by the invention can be used for improving the accuracy of an output noise level.

The invention provides a probability statistics method-based controller fatigue detection method. The method comprises the steps of obtaining brain waves of a controller, wherein the brain waves include slow alpha waves, alpha waves, beta waves and theta waves; performing calculation to obtain a power percentage of the slow alpha waves, a power ratio of the alpha waves to the beta waves, and a power ratio of the theta waves to the slow alpha waves; obtaining a probability statistics method-based fatigue detection model, and inputting the power percentage of the slow alpha waves, the power ratio of the alpha waves to the beta waves, and the power ratio of the theta waves to the slow alpha waves to the fatigue detection model, thereby obtaining a PERCLOS value simulation result; and performing controller fatigue detection according to the PERCLOS value simulation result. According to the probability statistics method-based controller fatigue detection method and system, brain wave parameters are input to the fatigue detection model built based on a probability statistics method to obtain the PERCLOS value simulation result, and the fatigue detection is performed; and the brain waves are used for indirectly reflecting fatigue levels of people, and a brain wave obtaining method is simple, easy to realize and low in cost.

The invention provides a selection method of optimal wavelet bases. The method comprises following steps: selecting wavelet bases with smallest average errors for reconstructed signals and noised signals to ensure precision of signals during the process of de-nosing wavelet thresholds; considering sparse characteristics of different wavelet bases in wavelet domains as for concrete characteristics of different signals, picking best wavelet basis for concrete characteristics of output signals in order to benefit de-nosing of wavelet thresholds. The overall method is easy and quick so that rapid and optimized selection of wavelet bases is achieved; and de-nosing effect and signal processing efficiency in engineering applications are improved.

The invention provides an edge detection method of a Canny operator based on high and low thresholds, which mainly comprises the following steps of: 1) carrying out smooth filtering processing on a source image, and removing noise of the image by using switch median filtering instead of Gaussian filtering; (2) calculating the gradient magnitude and direction of the image after smoothing filtering and denoising in the step (1) by adopting a sobel operator; 3) performing non-maximum suppression on the gradient magnitude obtained in the step 2) to obtain an edge image with a single-pixel width; 4) adopting a k-means clustering algorithm to obtain clustering centers of high and low gradient values; 5) obtaining an otsu threshold value of the gradient through an otsu algorithm, and taking a high threshold value and a low threshold value among the high clustering center, the otsu threshold value and the low clustering center; 6) processing the edge image of the single pixel width obtained in the step 3 by using a high threshold value and a low threshold value, and obtaining a binary edge; and 7) removing the interference edge of the binarized edge by adopting an area morphological opening operation, and obtaining a final edge image. The Canny algorithm provided by the invention has the advantages of high positioning precision, strong adaptability, good interference point removal effect and the like.

The invention relates to a fiber optic gyro (FOG) signal denoising method based on overlap M-band discrete wavelet transform (OMDWT). The FOG signal denoising method comprises the following steps of: establishing an FOG output signal model at first; carrying out OMDWT wavelet decomposition on an FOG output signal; then carrying out threshold processing on an OMDWT wavelet coefficient; and finally carrying out OMDWT wavelet reconstruction on the FOG output signal to obtain a denoised gyro signal. According to the invention, in the event of denoising the FOG output signal by utilizing the OMDWT, the wavelet coefficient and the scale coefficient of the OMDWT circularly move for corresponding beats along with circular motion of the FOG signal; in the conversion process, the gyro signal is decomposed according to a plurality of channels; the FOG signal denoising method has rapider decomposition speed according to different bands, more detailed band division on the gyro signal and a denoising effect prior to the traditional wavelet denoising effect; and the FOG signal denoising method disclosed by the invention has a important meaning for increasing the navigation performance of an inertial navigation system.

The invention relates to a spot image processing algorithm based on multi-scale wavelet transformation. The spot image processing algorithm includes the steps of firstly, performing logarithm transformation on an original image, and converting image multiplicative noise into additive noise; secondly, performing multi-scale wavelet decomposition on the image after the logarithm transformation; thirdly, selecting a threshold, and performing threshold treatment on a wavelet coefficient; fourthly, reconstructing the wavelet coefficient; fifthly, performing exponent operation to obtain the noise-reduced image. Compared with a traditional spatial filter noise reduction method, the spot image processing algorithm based on the multi-scale wavelet transformation has the advantages that the wavelet threshold noise reduction method has good visual effect, good noise reduction effect is achieved, and the edge information of the image can be kept effectively.

The present invention discloses a side channel attack preprocessing method based on a wavelet airspace correlation method. Collected power consumption signals f(n) are subjected to denoising through awavelet airspace correlation method and then is subjected to side channel analysis. According to the technical scheme, the interference of noise for a side channel attack result is effectively reduced, and the side channel attack success rate and the efficiency are improved.

The invention discloses a stepped mean filtering method aimed at the imaging data of multibeam forward-looking sonars, which belongs to the technical field of image processing. The imaging data of a single frame of sonar image are acquired, an imaging data matrix is generated, a stepped mask model is constructed, a mask coefficient is calculated, a data value corresponding to each point in the mask is determined, data convolution is carried out, stepped mean filtering is fulfilled, and the filtered imaging data matrix is visualized, so that a filtered multibeam forward-looking sonar image is obtained. Aiming at the imaging characteristics of a multibeam forward-looking sonar, the method provided by the invention can better inhibit noise, so that the filtered image has a better visual effect.

The invention discloses a cutter wear monitoring method for a numerically-controlled machine tool. The cutter wear monitoring method for the numerically-controlled machine tool comprises the followingsteps: acquiring a vibration acceleration signal and a microphone sound signal during working of a cutter of the machine tool as signals to be analyzed at first; pretreating an original signal by virtue of a rapid independent component analysis algorithm to obtain a vibration signal and a sound signal which are subjected to preliminary noise reduction, and then decomposing the acceleration signaland the sound signal by virtue of a variational mode decomposition algorithm to obtain respective intrinsic empirical mode functions separately; then carrying out denoising on existing modes by the rapid independent component analysis algorithm introducing a virtual noise channel to obtain effective mode components; adopting the maximum power spectrum density as a wear feature for extracting an intrinsic mode having the highest correlation degree with wear information; and finally composing a two-dimensional feature space by the wear features of the acceleration signal and the sound signal, and carrying out wear evaluation in a combined manner. The cutter wear monitoring method for the numerically-controlled machine tool is capable of effectively extracting the complete wear information,and realizing higher-stability and higher-resolution-ratio monitoring for the wear degree of the cutter.

The invention relates to a high-noise image denoising method based on a deep convolutional network. The method comprises the following steps: firstly, carrying out feature extraction on an image accommodating noise by adopting incremental expansion convolution, batch standardization operation and a Leakly ReLU function; then the image is recovered, and a mode of combining decreasing expansion convolution and a ReLU activation function is adopted; realizing separation of image noise and content by the network model through combination of residual learning and batch standardization operation; and finally, the optimal weight parameter of the network model is learned by solving the value of the minimized loss function (adopting different loss functions for different noise distributions). And finally, denoising the noise image by using the trained network model. The method can effectively remove the image noise in a high-noise environment, improves the visual effect of the image, and is better in practicality.

The invention discloses an image denoising method based on multi-resolution singular value decomposition. The method comprises the following steps of step1, adjusting a size of an original image, and carrying out singular value decomposition on an adjusted image data matrix to acquire a left singular matrix; step2, carrying out product on a transposed matrix of the left singular matrix and the adjusted image data matrix to acquire a new image data matrix; step3, carrying out size adjusting on each row of data of the new image data matrix; step4, taking a low frequency portion as an original image and repeatedly carrying out at least once process from the step1 to the step3; step5, using a threshold denoising rule to process a matrix corresponding to a high frequency portion, recombining a processing result into the image data matrix with the same size with the original image and acquiring a denoised image. In the invention, when the noise is removed, simultaneously, high frequency information of the image can be well kept, and a process is simple and is easy to carry out.

The invention discloses a double-residual denoising method based on an attention distribution mechanism, and the method comprises the steps: constructing a training data set, and carrying out the preprocessing of the training data set; constructing a network denoising model by using an attention distribution mechanism and a convolutional neural network of a double-residual network structure; setting a hyper-parameter and a loss function of the network denoising model, and optimizing the loss function; adding different levels of noise to the training data set and training to obtain a trained network model; image denoising is carried out on the trained network model, the noise image is evaluated through the structural similarity and the peak signal-to-noise ratio index, and the effect of improving the denoising performance and the imaging quality is achieved.

The invention discloses a signal-dependent noise parameter estimation method based on improved density peak clustering. A sample is extracted from an image containing noise through a sliding window, and a mean value, entropy and gradient are calculated to serve as feature data. And inputting into a clustering algorithm for clustering, and distinguishing weak texture samples from strong texture samples. In the clustering process, the concept of relative density is introduced, comparison ranges are divided through the distance between data points and surrounding data, the relative density of the data points in each comparison range is calculated, and finally the data points with high relative density are selected as clustering centers. The problem that when a traditional DPC algorithm is used for clustering a data set with non-uniform density, a sparse cluster center is often neglected, so that the clustering precision is influenced is solved. And according to a clustering result, pixel level estimation and noise estimation are carried out on a sample whose cluster label is weak texture, and finally a pixel value-noise variance estimation pair is fitted through a least square method, so that a noise parameter estimation value of an original image is obtained, and the preparation work of denoising is realized.

The invention discloses a full-bandwidth brain electrical signal obtaining device. The full-bandwidth brain electrical signal obtaining device is characterized in that an anti-shaking electrode comprises an electrode body, gel and an insulation layer, a direct current attenuator circuit only attenuates direct current and low frequency portions of an input signal, the amplitude of the direct current and low frequency portions is amplified and then is kept within the system A / D collecting amplitude value range, and the overflowing phenomenon does not appear; an amplification filter circuit comprises a secondary amplification circuit and a low pass filter circuit, and an upper computer restores the direct current portion through a digital compensation algorithm and is reconstructed with an alternating current portion to form an original brain electric signal. The full-bandwidth brain electrical signal obtaining device can meet the requirement for standard EDF format storage and has the advantages of being high in precision, full in bandwidth, strong in anti-jamming capacity, complete in data restoration and the like, and the direct current attenuation signal can be automatically reconstructed into the original full-bandwidth brain electrical signal under the conditions that the amplitude value range is from minus 100 mV to plus 100 mV and the precision is 0.3 mu V.

The present invention provides methods, systems, computer program products that use deep learning with neural networks to denoise ATAC-seq datasets. The methods, systems, and programs provide for increased efficiency, accuracy, and speed in identifying genomic sites of chromatin accessibility in a wide range of tissue and cell types.

The invention discloses a self-adaptive edge preserving denoising method based on an anisotropic diffusion model. The method comprises the following steps of preprocessing an original noise image, constructing a denoising algorithm model, performing iterative calculation on the original noise image and performing denoising processing on the original noise image. According to the method, the diffusion coefficient of the adaptive image denoising algorithm based on the combination of the fractional order differential operator and the Gaussian curvature is improved, bilateral filtering and local variance are added, the regularization item is introduced into the diffusion model, the image edge preserving effect is improved, the diffusion coefficient of the adaptive image denoising algorithm model is corrected, the denoising and edge maintaining effects are better, and the visual effect of the image is improved; the diffusion coefficient is adjusted by using the local variance so as to better control the diffusion speed; the image fidelity is improved by adding a regularization item, and an adaptive threshold is used, so that the medical image processing method is superior to a traditional image processing method in the aspect of processing a medical image besides a natural image.

The invention discloses an image processing method and device, a computer readable storage medium and a terminal. The image processing method comprises the following steps: acquiring a to-be-processed image; inputting the to-be-processed image to a trained neural network model, wherein the neural network model comprises a plurality of encoders and a plurality of decoders, different encoders output feature maps with different scales, each encoder fuses feature maps with different receptive fields or overlaps different color channels, and each decoder is used for recovering the encoded feature map and outputting a processed image; and outputting the processed image. According to the technical scheme, the image denoising and brightness improving effects can be improved.

The invention discloses a solar panel image processing method based on an adaptive algorithm, and the method comprises the steps: S1, obtaining a solar panel image, and converting the format of the solar panel image into RGB three channels; s2, carrying out image enhancement and grey-scale map conversion; s3, carrying out threshold binarization denoising on the solar panel image subjected to grey-scale map conversion; s4, performing morphological operation on the image; s5, setting contour detection constraint conditions, and performing contour detection to obtain the overall contour of the solar panel; s6, repeating the steps S2-S5 until the obtained contour number of the to-be-segmented region is greater than a preset value, and executing the step S7; s7, carrying out perspective transformation on the overall contour area of the solar panel to a predefined projection plane; and S8, carrying out image segmentation on the overall contour area of the solar panel according to the contourof each to-be-segmented area, and carrying out image zooming to obtain an image segmentation result. According to the method, efficient denoising during image segmentation is realized, and the methodhas relatively high expansibility and threshold adaptive effect stability.

The invention discloses a driveability evaluation test data processing wavelet denoising method. Compared with traditional wavelet denoising methods, the driveability evaluation test data processing wavelet denoising method comprises analyzing features of test data under different working conditions of automobile drivability testing, determining optimal wavelet base functions according to root-mean-square errors and signal-to-noise ratio, and according to the composite indexes structured by evenness and the root-mean-square errors, determining the optimal number of layers for wavelet decomposition. The driveability evaluation test data processing wavelet denoising method is high in good in driveability evaluation test data denoising effects and achieves guidance effects on data processingduring whole vehicle driveability test.

The invention discloses a medical PET image denoising method based on frequency domain direction smoothing Shearlet. A new medical PET image Gaussian noise model is provided, then frequency domain multi-scale decomposition and multi-smooth direction decomposition are carried out, new unified threshold processing is carried out on the decomposed high-frequency direction smooth Shearlet coefficient,and then a denoised PET image is generated through inverse Shearlet transformation. Compared with a traditional NSST method (non-subsampled Shearlet transform), the method has the advantages that thedenoising effect is better, the speed is higher, and the method can be better applied to the field of medical PET image denoising.

The invention discloses a method for denoising based on a pulse type ToF depth image. The method comprises the following steps: directly starting from an imaging model of a pulse type ToF depth camera, designing a regularization device of original measurement data of a pulse type ToF sensor, carrying out depth denoising by adopting a variation-based semi-blind deconvolution algorithm, effectivelydenoising a pulse type ToF depth image, and keeping edge detail information of the depth image.

The invention relates to the technical field of oil-gas exploration, in particular to a fault enhancement method, a fault development interpretation method, a storage medium and electronic equipment. The problems that in the prior art, a method for conducting fault enhancement on seismic data with poor quality and weak fault information is large in calculated amount, the enhancement effect of small faults is not obvious, and poor fault continuity is difficult to overcome are solved. The method comprises the steps of obtaining three-dimensional seismic data of a target layer section, calculating an edge detection body and performing fault enhancement processing to obtain a fault enhancement body; variance operation and ant tracking operation are sequentially carried out on the fault enhancement body, angle control is carried out according to the fault development direction, continuity enhancement processing is carried out on the fault body in the fault development direction, and a three-dimensional seismic attribute body of the target layer section after feature enhancement is obtained; the purposes of improving the signal-to-noise ratio and the imaging quality of the seismic data, further depicting the fault clearly and facilitating subsequent fault recognition and extraction are achieved.