14 results about "Total variation model" patented technology

The invention discloses an escalator passenger abnormal behavior detection method and device, a medium and equipment, and relates to the technical field of abnormal behavior detection.The escalator passenger abnormal behavior detection method comprises the steps that an obtained to-be-detected video stream is divided into continuous frames, and an initial three-dimensional tensor used for representing passenger behavior information is obtained according to the continuous frames; constructing a joint optimization model based on a tensor ring low-rank model TRLRD and a tensor total variation model TTV; decomposing the initial three-dimensional tensor through a joint optimization model, determining a low-rank background and a dynamic foreground, and separating the low-rank background from the dynamic foreground to obtain a target foreground tensor; and inputting the target foreground tensor into a multi-person attitude estimation algorithm AlphaPose, performing attitude estimation on the passengers, determining an attitude detection result, inputting the attitude detection result into a trained convolutional neural network for abnormal behavior recognition, and obtaining a recognition result.

The invention discloses a moving target detection method based on low-rank full-connection tensor network decomposition and a total variation model, and belongs to the technical field of computer vision and real-time video analysis. Aiming at the problems of poor real-time performance, dynamic background misjudgment, serious noise interference and the like caused by neglect of space-time relevance, fixed rank constraint and high calculation complexity of a traditional method in a complex dynamic scene, the following technical scheme is provided: through full-connection tensor network (FCTN) decomposition, a common rank is utilized to realize accurate modeling of a dynamic background; interference from background leakage to foreground is reduced; incremental updating of time core parameters is carried out in combination with a sliding window strategy, and a historical space core is fixed to reduce calculation complexity; total variation (TV) regularization is combined to optimize a low-rank-sparse separation process, the continuity of a space gradient and a time gradient is restrained, the contour integrity of a moving target is enhanced, and noise is suppressed.

The application provides a low-quality welding image enhancement method for preserving structure and balancing brightness, and the specific steps are as follows: inputting a low-quality welding original image, forming an initial illumination component image output through a principal component analysis algorithm; processing the initial illumination component image by using a relative total variation model to form an illumination component image output with optimized structure characteristics; dividing the illumination component image with optimized structure characteristics into two paths, one path is directly substituted into a Retinex model together with the original image to obtain a reflection component image, and the other path is processed by a logarithmic transformation enhancement algorithm to obtain an illumination component image with optimized brightness; fusing the reflection component image and the illumination component image with optimized brightness according to the Retinex model to obtain a reconstructed image; and introducing a guide filter to take the initial illumination component image as a guide image to obtain a final enhanced image. The image enhanced by the method has uniform and natural brightness distribution, good structure characteristic information preservation, no artifact or block effect, and clearer image details.

The application discloses an X-ray scanning imaging method based on an X-ray flat panel array, relates to the technical field of scanning imaging, and breaks through the traditional dose constraint by simulating high dose noise characteristics in a projection domain based on virtual dose enhancement; based on dynamic registration of bone probability weighting, the method eliminates the blurred edge of a vertebral body caused by respiratory displacement without the need of external sensors or patient cooperation; by means of an iterative weight shrinkage mechanism, a weighted total variation model can inhibit soft tissue noise while preserving micron-level features such as bone trabeculae, and over-smoothing distortion is avoided; frame frequency control in a respiratory phase is self-adaptive, compatible with existing flat hardware, and can directly upgrade existing equipment, thereby reducing hospital procurement costs.

The invention discloses an X-ray scanning imaging method based on an X-ray flat plate array, relates to the technical field of scanning imaging, and aims to simulate high-dose noise characteristics in a projection domain by virtual dose enhancement and break through traditional dose constraints; based on dynamic registration of bone probability weighting, ambiguity of the vertebral body edge caused by respiratory displacement is eliminated, and cooperation of an external sensor or a patient is not needed; the weighted total variation model retains micron-sized features such as bone trabecula and the like while inhibiting soft tissue noise through an iterative weight contraction mechanism, and excessive smooth distortion is avoided; respiratory phase self-adaptive frame frequency control is compatible with existing panel hardware, existing equipment can be directly upgraded, and the purchase cost of hospitals is reduced.

The application discloses a SAR and optical image fusion method and system based on morphological feature decomposition, and belongs to the field of image processing. First, the optical image and the SAR image are preprocessed, then, a differentiated multi-layer feature decomposition strategy is adopted to decompose the image into main structure information, morphological feature information and edge detail information, wherein the main structure information is extracted by a local total variation method, the morphological feature information is obtained by a generalized total variation model and residual suppression, and the edge detail information is decomposed by jointing bilateral filtering and a least square model. Then, different feature layers are respectively fused by a saliency guided fusion, a local weighted fusion and a fusion strategy combining edge preservation and sparse detail compensation. Finally, color information is restored by IHS inverse transformation to obtain a fused image with clear structure, complete morphology and rich details, so that fine fusion from structure to morphology and then to details is realized.

The application provides a multi-focus image fusion method combining a deep residual network and a variational method, and relates to the technical field of digital image processing. The application firstly acquires a first source image and a second source image which are complementary to a focusing area, determines a first-order gradient and a first-order gradient module of the first source image and the second source image; outputs a focus point image and a gradient module score image through an MResNet deep residual network after training, generates an initial fusion image and a fusion gradient guide term; constructs a light-weighted total variation model which only adopts a two-direction first-order gradient operator, and obtains a total variation fusion result at a junction; and performs block fusion based on a weight map to obtain a final fusion image. The application can retain clear area information of source images and improve the transition continuity at a focusing and defocusing junction.

The application belongs to the technical field of target detection, and relates to an infrared dim small target detection method, device, equipment and medium. The method comprises the following steps: acquiring an infrared sequence image and obtaining a space-time tensor model; constructing a three-dimensional structure tensor according to the infrared sequence image to obtain a background edge information tensor; calculating a time-constrained Gaussian curvature filter according to the infrared sequence image to obtain a filter feature map tensor; obtaining a feature weight map according to the background edge information tensor and the filter feature map tensor; constructing a weighted space-time total variation model and defining a non-convex function; constructing an augmented Lagrange function according to the space-time tensor model, the feature weight map, the weighted space-time total variation model and the non-convex function; solving the augmented Lagrange function by using an alternating direction multiplier method until convergence is achieved, to obtain a background tensor and a target tensor, and obtaining an infrared dim small target from the target tensor. The application can accurately detect the infrared dim small target.

The invention relates to the field of image processing, in particular to an infrared and visible light image fusion method based on a symbiotic statistics induced weighted total variation model, which comprises the following steps of: performing gray level quantization on an infrared image, and constructing a local gray level symbiotic matrix for each pixel point to calculate a symbiotic probability; generating a spatial adaptive symbiotic weight map capable of quantifying the regional texture flatness based on the probability; introducing the weight map into a total variation regular term, constructing a weighted total variation model, and adaptively adjusting the smoothness intensity of different regions of the image by the model by using the weight map; and solving the model by adopting an alternating direction multiplier method to obtain a final fusion image. According to the method, adaptive noise and edge distinguishing is realized, and the thermal noise of the infrared image is effectively suppressed. According to the method, adaptive noise and edge distinguishing is realized through symbiotic statistics, thermal noise of the infrared image is effectively suppressed, detail textures of the visible light image are reserved, and the visual quality and information integrity of the fused image are effectively improved.

The application provides a mixed noise hyperspectral image restoration method combining double low rank and empty spectrum total variation, which uses low rank tensor approximation and low rank matrix approximation model of each band respectively to excavate low rank properties of noise-free hyperspectral remote sensing image and strip noise, simultaneously introduces anisotropic spatial spectrum total variation model, establishes a hyperspectral remote sensing image multi-type mixed noise removal model combining double low rank approximation and anisotropic spatial spectrum total variation, and uses an alternating direction multiplier method to obtain noise-free hyperspectral remote sensing image. The application is applied to mixed noise removal of hyperspectral remote sensing image of Gaofen 5, and real hyperspectral remote sensing image experiments of Gaofen 5 show that the application can more effectively remove multi-type, high-intensity mixed noise in the hyperspectral remote sensing image of Gaofen 5, simultaneously protects high-dimensional structure information of the image, and greatly improves the quality of the hyperspectral remote sensing image of Gaofen 5.

This invention relates to the field of digital image processing technology and is an adaptive optics retinal image deblurring method based on a non-convex total variation model. The method includes: constructing an adaptive optics retinal image myopia deconvolution model; improving the adaptive optics retinal image myopia deconvolution model based on norm-based non-convex total variation regularization; constructing a constrained optimization model equivalent to the improved adaptive optics retinal image myopia deconvolution model; using the alternating direction multiplier method as the outer optimization algorithm to estimate the adaptive optics retinal image and the weights to be estimated; solving the subproblems in the alternating direction multiplier method; updating the Lagrange operator and calculating the stopping condition according to a given tolerance to obtain the updated adaptive optics retinal image myopia deconvolution model; and using the updated adaptive optics retinal image myopia deconvolution model to perform the retinal image deblurring operation.

The invention discloses a multi-modal remote sensing image registration method based on significant structure feature guidance, and relates to the technical field of image registration. The method comprises the following steps: firstly, obtaining a multi-scale structure diagram based on a multi-scale multi-direction relative total variation model, and extracting edge features in different directions by using a first-order Gaussian direction adjustable filter; gathering edge features, counting maximum response amplitudes of pixels in all directions to generate an edge image, and performing weighted fusion on the edge image and phase consistency maximum moment information to construct a significant edge image; constructing an edge index map based on the index of the maximum response amplitude, and obtaining a salient edge index map through salient edge map guide optimization; and finally, determining a main direction index on the significant edge index map through a global information rotation strategy, constructing a feature descriptor, and finishing feature matching in combination with a nearest neighbor distance ratio and a rapid sampling consistency method. According to the method, images with obvious nonlinear gray scale, rotation difference and small-scale change can be successfully registered, and the method is suitable for an image matching task in a complex scene.

The present application relates to the field of image processing, in particular to an infrared and visible light image fusion method based on a co-occurrence statistics induced weighted total variation model, the gray level of an infrared image is quantized, and a local gray co-occurrence matrix is constructed for each pixel point to calculate a co-occurrence probability; a spatial adaptive co-occurrence weight map capable of quantifying the texture flatness of a region is generated based on the probability; the weight map is introduced into a total variation regular term to construct a weighted total variation model, the model adaptively adjusts the smoothing strength of different regions of an image by using the weight map; the model is solved by using an alternating direction multiplier method to obtain a final fusion image. The present application realizes adaptive region differentiation of noise and edges, and effectively suppresses thermal noise of an infrared image. The present application realizes adaptive region differentiation of noise and edges by using co-occurrence statistics, effectively suppresses thermal noise of an infrared image, and retains the detail texture of a visible light image, effectively improving the visual quality and information integrity of a fusion image.

The present application relates to a remote sensing image non-uniformity correction method based on neighborhood statistics and total variation model, and belongs to the technical field of image processing. The present application is aimed at the abnormal pixel response problem existing in the imaging process of a remote sensing camera and the non-linear response problem caused by factors such as manufacturing process, and sequentially performs abnormal pixel and non-uniformity correction processing on the imaging data of the remote sensing camera. The abnormal pixel correction is completed by using the neighborhood information of the abnormal pixels in the remote sensing image, the non-uniformity correction is based on the total variation model to solve the optimal correction result of the remote sensing image, and the texture details of the image are effectively preserved in combination with the adaptive wavelet transform. The remote sensing camera non-uniformity correction algorithm proposed by the present application can effectively correct the abnormal pixels and non-linear response in the remote sensing image.