1589 results about "Image denoising" patented technology

The invention provides an image denoising method based on generative adversarial networks, and belongs to the technical field of computer vision. The method comprises the following steps: (1) designing a neural network for estimation for noise intensity of an image containing noises; (2) using image blocks in an image library to add noises of the intensity according to the estimated noise intensity to use the same as samples of training the networks; (3) in network training, designing a new generation network and discrimination network, and adopting a form of fixing the generation network to train the discrimination network and fixing discrimination network parameters to train the generation network to enable the networks to carry out adversarial training; and (4) using the trained generation network as a denoising network, and selecting a network parameter according to a result, which is obtained by the noise recognition network, to denoise the image containing the noises. The methodhas the effects and the advantages that a visual effect of the denoised image is improved without the need for manual intervention for adjusting the parameter, and texture details of the image can bebetter restored.

An “Image Denoiser” provides a probabilistic process for denoising color images by segmenting an input image into regions, estimating statistics within each region, and then estimating a clean (or denoised) image using a probabilistic model of image formation. In one embodiment, estimated blur between each region is used to reduce artificial sharpening of region boundaries resulting from denoising the input image. In further embodiments, the estimated blur is used for additional purposes, including sharpening edges between one or more regions, and selectively blurring or sharpening one or more specific regions of the image (i.e., “selective focus”) while maintaining the original blurring between the various regions.

The invention discloses a color digital image enhancing and denoising method under random illumination. The method is characterized by comprising the following steps of: 1, image denoising processing, namely converting an original image from an RGB space to a YCbCr space, removing a Gaussian noise by using a Gaussian filter and removing a salt and pepper noise by performing median filter; 2, image brightness / contract stretch processing, converting the image from the RGB space to an HSI space, decomposing the image by using a two-sided filter, processing by an improved Retinex model algorithm, and obtaining a new image saturation by performing saturation compensation; and 3, performing fusion display on the image acquired after denoising the YCbCr space and the image acquired by HSI space processing. By the method, the color constancy of the image can be kept, the dynamic range of the image can be improved well, and simultaneously noises of the image can be inhibited and removed with less texture and detail information loss.

An “Image Denoiser” provides a probabilistic process for denoising color images by segmenting an input image into regions, estimating statistics within each region, and then estimating a clean (or denoised) image using a probabilistic model of image formation. In one embodiment, estimated blur between each region is used to reduce artificial sharpening of region boundaries resulting from denoising the input image. In further embodiments, the estimated blur is used for additional purposes, including sharpening edges between one or more regions, and selectively blurring or sharpening one or more specific regions of the image (i.e., “selective focus”) while maintaining the original blurring between the various regions.

The invention discloses an ReLU convolutional neutral network-based image denoising method. The method comprises the following steps of building an ReLU convolutional neutral network model, wherein the ReLU convolutional neutral network model comprises a plurality of convolutional layers and active layers after the convolutional layers, wherein the active layers are ReLU functions; selecting a training set, and setting training parameters of the ReLU convolutional neutral network model; training the ReLU convolutional neutral network model by taking a minimized loss function as a target according to the ReLU convolutional neutral network model and the training parameters of the ReLU convolutional neutral network model to form an image denoising neural network model; and inputting a to-be-processed image to the image denoising neural network model, and outputting a denoised image. According to the ReLU convolutional neutral network-based image denoising method disclosed by the invention, the learning ability of the neural network is greatly enhanced, accurate mapping from noisy images to clean images is established, and real-time denoising can be realized.

An Active Random Field is presented, in which a Markov Random Field (MRF) or a Conditional Random Field (CRF) is trained together with a fast inference algorithm using pairs of input and desired output and a benchmark error measure.

The invention discloses a convolutional neural network medical CT image denoising method based on residual error learning. The method comprises the following specific steps: 1) constructing a medicalCT image model; 2) constructing a neural network model; 3) training the network; 4) updating parameters; and 5) denoising the medical CT image: inputting the medical CT image containing the noise intothe constructed network model, and outputting the medical CT image without the noise through the network. The method has the advantages that medical CT image denoising is carried out by combining knowledge in the aspect of the convolutional neural network in deep learning. Noise in the image is approximately learned in a residual error learning mode, good pertinence is achieved, and meanwhile thetraining efficiency of the neural network is improved. The convolutional neural network and the residual error learning method are adopted, so that the feature information in the image can be betterlearned, and more local image information can be reserved in the image denoising process. And meanwhile, the image denoising capability is also improved.

The invention provides a QR recognizing method based on image processing. The method includes the following steps of (1) preprocessing which includes graying, image enhancement, binaryzation and image denoising; (2) locating which includes determining position information of 12 angle points on a QR code sign position detection graph and position information of angle points at the right lower corner of the QR code symbol; (3) decoding which includes conducting geometrical correction according to various position information of the acquired QR code signs in the locating phase, extracting standard QR code signs and drawing a sampling network. The method has the advantage that the method can be applied to the automatic field of industrial production, achieves an automatic recognition function of QR codes without manual participation, and can solve the problem in actual application that the image background is complex, the image quality is not high, and the QR code position detection image proportional characteristics are damaged. The algorithm can reach the high identification rate of 98% under the stable operation condition. The algorithm is high in real-time performance. Time used by recognizing a frame of an image is about 150 ms.

The invention discloses a cloth defect detection method based on machine learning. The method particularly comprises the following steps of image segmentation, image enhancement, image denoising, sub image feature extraction, defect point area segmentation, offline cloth learning, online cloth detection, and cloth defect point classification. In the offline cloth learning stage, a BP neural network is used to train standard image characteristic parameters, and a standard value is obtained. In the online cloth detection stage, the BP neural network is used to detect the feature parameters of the sub image. In the cloth defect point classification stage, a depth learning algorithm based on a convolutional neural network is used to classify cloth defects. The cloth defect detection method based on machine learning has a self-learning function and can meet the continuously-developing industry needs. The invention also provides a cloth defect detection device based on machine learning, which comprises an image acquisition unit, an image processing unit, a data communication unit and an action execution unit. The detection device can realize high-efficiency and accurate detection, and workers can be freed from heavy and tasteless physical labor.

The invention relates to an SAR image denoising method based on a multi-scale hole residual attention network. The method comprises the following steps of by extracting features of different scales ofthe image through a multi-scale convolution group, broadening a convolution kernel receptive field by utilizing cavity convolution; exuecting more context information of the image; and transmitting the feature information of the shallow layer to a deep convolutional layer by using jump connection to keep image details, adding an attention mechanism to intensively extract noise-related features, and automatically learning the SAR image speckle noise distribution form in combination with a residual error learning strategy to achieve the purpose of removing speckle noise. Experimental results show that compared with a traditional SAR image noise removal method, the method has the advantages that the speckle noise removal effect is good, the number of artificial traces is small, detail information of the image is kept, and the calculation speed is higher through the GPU.

The invention discloses an image denoising method based on a generative adversarial network, and the method is characterized by comprising the following steps: 1, selecting an experiment data set; 2,selecting Gaussian additive white noise as a noise model; 3, building a generative network model, and training a generator network G for denoising; 4, establishing a discrimination network model, wherein a discriminator D is used for carrying out authenticity classification on the input image; 5, constructing a joint loss function model; 6, training a generative adversarial network; and step 7, carrying out image denoising quality evaluation. According to the image denoising method based on the generative adversarial network, the denoising effect of reserving more texture details and edge features can be achieved.

The present invention discloses a depth image denoising and enhancing method based on deep learning. The method comprises the steps of establishing a depth image denoising and enhancing convolutional neural network, wherein the network is composed of three layers of convolution units which finish the functions of feature extraction, non-linear mapping and image reconstruction of the input images respectively; jointly using the depth and visual images as the input of the convolutional neural network, wherein firstly the visual images are processed into the grayscale images in a grayscale processing manner; and enhancing the edge information and taking out the redundant information by the image preprocessing; segmenting the depth images into the image blocks according to certain intervals, adding the effective data by a rotation and pixel overturning data amplification method, and discarding the interference blocks and the redundant blocks; and improving the learning efficiency of the network adaptively based on a loss training depth image enhancement convolutional neural network of a weight map. According to the method of the present invention, the black spot filling and the denoising operations can be carried out on the depth images with noise real-timely, and the good visual effect and depth value recovery effect can be realized.

The invention discloses a Camare360 mobile phone photographic platform and an instant processing method. The Camare360 mobile phone photographic platform comprises a main interface module, a photographic effect module, an image processing module, an optical lens simulation module, an anti-vibration module, a segmental shrinking developing instant image processing module and an internet service module. A method for instantly processing a photo comprises the following steps of: executing a shutter button module; extracting and analyzing image data; and executing a multi-layer feedback queue algorithm, a high-speed image scaling algorithm, an MHDR fast algorithm, a simulated optical lens scattered aperture visual algorithm, an ABM3D-based image denoising algorithm and a high-speed color mixing algorithm. By a platform for instantly processing the photo by using a third generation (3G) mobile phone, the quality of mobile phone photos is greatly improved; artistic representation and artistic impact of mobile phone photography are greatly enriched; the complex processes are simplified to be integrated fully automatic operation; and a user can take various types of high-quality artistic photos as the user wishes.

The invention relates to the field of video image processing and discloses motion-adaptive video image denoising method and device. Time-domain denoising strength is controlled by determining motion level of a current frame to be processed, and noise level of the frame, to be processed, subjected to time-domain filtering is updated. The frame to be processed is subjected to space-domain filtering according to the updated noise level. Therefore, strength in time-domain denoising and space-domain denoising is adaptively adjusted. When the motion level decreases, the time-domain denoising strengthens and the space-domain denoising weakens; on the contrary, the time-domain denoising strength decreases and the space-time denoising strength increases; therefore, better denoising is achieved, the problem that video-denoised images blur and motion objects are 'tailed' are well solved, and system implementation cost can be effectively controlled.

The present invention discloses a hyperspectral image denoising method based on non-convex low rank matrix decomposition. The method comprises the following steps: 1. dividing a hyperspectral image; 2. generating a matrix; 3. carrying out nonconvex low rank matrix decomposition; 4. restoring the sub-block; and 5. restoring the hyperspectral image, and finally obtaining the denoised hyperspectral image. According to the method disclosed by the present invention, a method for efficiently and quickly removing multiple kinds of noises is provided for the remote sensing hyperspectral image denoising.

The invention discloses an intelligent identification and analysis method for an examination answer sheet system. The method comprises the following steps: carrying out item-by-item decomposition operation on image gray scale, image binaryzation, image denoising and morphological processing in an image preprocessing technology; an independent peak point is formed by using a monitoring image inclination and rotation correction technology, and an image is read for binaryzation; Edge monitoring operator monitoring is constructed by using an edge monitoring technology and through an original imagedifferential technology, and image filtering, image enhancement, image detection, a Canny operator and a Canny edge recognition algorithm are formed; and data image recognition analysis is carried out by adopting image recognition point positioning and affine transformation. The method can achieve the intelligent high-speed recognition of the answer sheet, is higher in image precision, is high inefficiency, is not limited to specific scanning equipment, achieves the mining and analysis of large data of batch scanning results, can intelligently recognize specific wrong knowledge points, and can achieve the intelligent statistical induction.

There is provided a method of reducing block noise, mosquito noise and other image noises in an image by a filtering process using a smoothing filter, which block noise, mosquito noise and other image noises being caused at the time of decoding encoded, compressed image data on a block-by-block basis. The method includes changing the extent or intensity of the image noise reduction in the filtering process in a continuous or stepwise manner according to an output size or expansion rate of an image to be outputted to printer paper, photographic paper or other output media, thereby allowing the extent or intensity of the image noise reduction to be increased as the output size or expansion rate of the image increases. This method is capable of allowing the extent of the image noise reduction applied to image data to be perceived in a similar fashion, irrespective of the output size.

The invention relates to the field of image processing, and provides a novel self-adaptive secondary noise point detection method based on direction information. The method can be used for effectively reducing the erroneous judgment probability of non-noise points and removing impulse noise in images more effectively, and the denoising robustness on noise with different strengths is stronger. To this end, the technical scheme adopted in the invention is as follows: an image denoising method based on secondary noise point detection comprises the following steps: at first, determining an impulse noise density condition, counting all gray values in an image, calculating the number of pixel points whose gray values are 0 or 255, dividing the number by the number of total pixel points to obtain an approximate noise probability density pa, when pa is larger than or equal to s, filtering all points whose gray values are 0 or 255, when pa is smaller than s, carrying out secondary noise point detection on these points, and the detection principle is mainly based on direction features of edge information in the image; and then self-adaptive filtering is carried out. The image denoising method based on secondary noise point detection provided by the invention is mainly applied to image processing occasions.

The invention discloses an image denoising method based on a neural network. The image denoising method based on the neural network includes the steps of acquiring high-sensitivity images and low-sensitivity images of a scene, generating a first neural network model, denoising the low-sensitivity images via the first neural network model to obtain denoised low-sensitivity images, constituting training image database through the high-sensitivity images and the denoised low-sensitivity images and using the training image database to train a neural network model so as to generate a second neuralnetwork model which is used for image denoising, using images to be processed as input to the second neural network model and then acquiring corresponding clean images after treatment. Furthermore, aimage denoising system based on the neural network is also disclosed. The above-mentioned image denoising method and system based on the neural network have the advantages of being capble of processing images with real noise information and having practical application value.

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.

This invention is related to a cognitive Multi-Input Multi-Output (MIMO) radio frequency (or laser) radar with large-area synchronized multi-dimensional hopping spread spectrum and Interference-Free Windows (IFWs) for autonomous vehicles comprising of (1) analog component; (2) digital baseband component; (3) multi-dimensional hopping code generator; (4) large-area time synchronization; (5) cooperative IFW; and (6) cognitive engine. The new MIMO radar can provide interference-free environmental perception intelligently by the multi-dimensional IFWs formulated by beat-frequency hopping, beat-time hopping, discrete sequence (DS), cooperative IFW, array processing, and image denoising and fusion, etc. This invention increases the frequency efficiency greatly, and be applied to autonomous vehicles and robotics under sparse, dense, mixed autonomous and human driving, or completely autonomous driving environments.

The invention provides a synthetic aperture radar (SAR) image bionic recognition method based on sample generation and nuclear local feature fusion and belongs to the field of image processing technologies and SAR target recognition. According to the method, a super complete training sample set is firstly constructed for training to obtain geometry manifold, then a sample to be recognized is recognized, specifically, each sample is firstly subjected to image denoising by a K-SVD dictionary learning method, and object region extraction is achieved by means of an object centroid method; and feature extraction is performed respectively by combining local phase quantization (LPQ) and a Gabor filtering method, feature fusion is performed, finally, classification is performed by covering of high-dimensional geometry manifold, and recognition is performed by a bionic mode. According to the SAR image bionic recognition method based on sample generation and nuclear local feature fusion, inhibiting effects of image coherent noises are obvious, SAR image features can be effectively extracted, the problem of the unstable extracted features, which is caused by changes of attitude angles of SAR images, is solved, the recognition accuracy is high, and the method has good robustness.

The invention discloses a non-local image denoising method based on low rank restoration. The method comprises the steps of (1) collecting an image, (2) carrying out gray transformation on the collected image, (3) establishing a three-dimensional similarity matrix through the global search of a similar pixel block for the image which is subjected to gray transformation, then carrying out hard threshold filtering on the discrete cosine transform and Hadamard transform coefficients of the three-dimensional matrix is carried out, obtaining the initial estimation of the similar pixel block with the removal of partial noise, and improving the matching accuracy of the similar pixel block with low rank restoration denoising while removing a large part of noise, and (4) with transform domain filtering as prior knowledge, searching a similar pixel block in a search area for an initially denoised reference pixel block, then forming the similarity matrix by using a similarity block corresponding to an original image, carrying out low rank matrix decomposition on the similarity matrix, effectively separating noise and a signal, and obtaining a final denoised image. The method has the advantages of a simple and fast algorithm, a high signal-to-noise ratio and good consistency and is especially suitable for the requirements of high quality noise reduction of large-scale images.

Systems and methods for multispectral imaging are disclosed. The multispectral imaging system can include a near infrared (NIR) imaging sensor and a visible imaging sensor. The disclosed systems and methods can be implemented to de-noise a visible light image using a gradient scale map generated from gradient vectors in the visible light image and a NIR image. The gradient scale map may be used to determine the amount of de-noising guidance applied from the NIR image to the visible light image on a pixel-by-pixel basis.

The invention discloses a rotating arc narrow gap MAG (metal active gas) welding seam offset identification device and method based on visual sensing. The device comprises a rotating arc welding torch, an arc rotational position sensor, an insulation module, a CCD (charge coupled device) camera and a PC (personal computer). The PC comprises an image capture card, an image denoising module, an arc center recognizing module, a bevel edge recognizing module, a current arc rotating period welding seam offset extracting module, and a counting module. The front of the CCD camera is provided with dimmer glass, an optical filter and a UV lens. The method includes adopting the CCD camera and triggering a sampling manner to accurately acquire welding images that an arc rotates to the left and right of a bevel in each arc rotating period, and rapidly detecting the left and right bevel positions of the arc in the two images to acquire welding seam offset. According to the device and the method, efficiency of acquiring the welding seam offset is high, calculating method is accurate and reliable, and bevel bottom changes and arc electrical signals are affected rarely, and the method and the device are adaptable to high-frequency rotating arc welding.

The invention proposes a hyperspectral image denoising method based on robust low-rank tensor, which comprises the steps of establishing a mathematical model of hyperspectral image noise, constructing a hyperspectral image robust low-rank tensor (RLRTR) denoising optimization model, and solving the RLRTR denoising optimization model. The hyperspectral image denoising method fully utilizes the prior knowledge of hyperspectral images (HSI) which are polluted by different kinds of noise such as Gaussian noise, impulse noise, dead pixels and striping noise. The hyperspectral image denoising method utilizes the potential low-rank tensor property of clean hyperspectral image data and the sparsity property of abnormal and non-Gaussian noise, and adopts a nuclear norm and an l2,1 norm for representing the low-rank and sparsity properties. The technical scheme of the hyperspectral image denoising method fully uses the prior information and internal structure characteristics of the hyperspectral image, and can remove Gaussian noise, abnormal and non-Gaussian noise simultaneously.

The invention relates to an image denoising processing method and apparatus. The method comprises the steps of receiving a shooting instruction; performing continuous shooting according to the shooting instruction to obtain multiple frames of original images; obtaining an environmental brightness value and / or an ISO value corresponding to a shooting environment; judging whether one of at least following conditions is met: the environmental brightness value is smaller than or equal to a preset brightness value, and the ISO value is greater than or equal to a preset ISO value; if yes, obtaining a to-be-denoised original image according to the environmental brightness value or the ISO value; and obtaining RAW data of the to-be-denoised original image, converting the RAW data into YUV data, and performing multi-frame denoising processing on the YUV data to obtain an output image. With the adoption of the method, a clear image can be obtained without manually adjusting exposure time or brightness gain in an automatic shooting mode.

The invention discloses an image denoising method. The method comprises steps: an average block set of a to-be-denoised image is acquired; a corresponding mean value is subtracted from each average block in the average block set to obtain a noise block set; a first noise image in the noise block set is acquired; noise modeling is carried out on the noise block set according to a generative adversarial network to obtain a generator capable of generating the same kind of noise as the to-be-denoised image; a second noise image is acquired according to the generator; according to a noise-free image, the first noise image and the second noise image, a training set is constructed; according to the training set and a discriminative learning method, an image denoising network model is trained; andthe to-be-denoised image is inputted to the image denoising network model to acquire an image after denoising. The invention also provides an image denoising device, equipment and a storage medium. the denoising effects on unknown real noise in real life can be enhanced and the denoising efficiency is improved.

The invention relates to an improved threshold function-based wavelet transformation image denoising method which comprises the following steps of: firstly selecting a base wavelet, determining the number of wavelet decomposition layers and carrying out multi-scale wavelet transformation on noise-containing images; respectively determining a threshold for each layer of detail coefficient; carrying out threshold treatment by using improved threshold functions; and finally carrying out wavelet reconstruction on low-frequency coefficients and threshold treated high-frequency coefficients to obtain the denoised images. According to the denoising method provided by the invention, the defect of discontinuity of the hard threshold functions is overcome and the constant deviation in the soft threshold functions is decreased at the same time. By using the denoising method, the signals and the noises can be distinguished effectively, the image information edges are protected when the noises are removed, the peak signal to noise ratio of the images is improved, and better image denoising effect is achieved.

A method and an apparatus are disclosed by which both random and systematic noise may be reduced in electronic images.