185 results about "Structure tensor" patented technology

The invention provides a video real-time de-noising method based on a structure tensor and Kalman filtering. The method includes the following steps that an image frame to be processed at the current moment is acquired, and n frames of images which are before the current frame and have been stored and de-noised; the image frame to be processed at the current moment is pre-filtered with an average filter; based on the structure tensor, movement of the current image frame is estimated by fully utilizing close time and space relations between the image frame to be processed at the current moment and adjacent image frames before the image frame to be processed at the current moment; based on the movement estimation result, de-noising is performed in a time domain with a Kalman filtering method; de-noising is performed in a space domain with a Wiener filtering method; two de-noised images are synthesized, and the final de-noised image is acquired through weighting. With the method, high-noise video can be de-noised, and a good de-noising effect is achieved; besides, because complicated interactive calculation does not exist, achievement of hardware such as FPGAs is facilitated, and real-time the high-noise video can be de-noised.

A method is disclosed for enhancing the quality of an image dataset, such as, e.g., reducing the noise in a noisy image data set and increasing the contrast in the image data set. The method may be used for processing a sequence of image datasets, e.g. night vision image datasets, wherein said sequence comprises at least two image datasets each having at least two pixels, and wherein each pixel has an intensity value. The method comprises calculating a structure tensor for each pixel in an image dataset comprised in the sequence of image datasets; calculating values in a summation kernel based on said structure tensor for each pixel in said image dataset; calculating a weighted intensity value for each pixel in said first image dataset, using as weights the values in said summation kernel; storing said weighted intensity value for each pixel in said image dataset as a processed intensity value for each corresponding pixel in a processed output image dataset; rotating a local coordinate system in which the summation kernel is described resulting in that the coordinate axes of said local coordinate system coincide with the directions of the eigenvectors of said structure tensor, where said eigenvectors are described in the global coordinate system of the image dataset, and scaling the coordinate axes of the local coordinate system in which the summation kernel is described by an amount related to the eigenvalues of the structure tensor via a width function W(λ_i)=σ_i, and wherein said eigenvalues depend on the amount of intensity variation in the direction of their corresponding eigenvectors, the width function being a decreasing function such that w(0)=σ_max and lim_a→∞w=σ_min. An apparatus and a computer readable medium are also provided.

The invention relates to a method and a device for extracting characteristic points of a lane line under a complex road condition. The method comprises the following steps of: acquiring a grayscale image of a road by using vehicular equipment; dividing the grayscale image into an irrespective region and a region of interest (ROI) according to the degree of importance of the road image to the characteristic point extraction of the lane line; performing smoothing filtering processing on the ROI; and performing binarization processing on the smoothing-filtering processed ROI by utilizing a structure tensor field-based level set extrinsic curvature, and extracting the characteristic points of the lane line. The problem of low robustness for the characteristic point extraction of the lane line under the complex road condition such as illumination variation, shadowing, breakage and smear coverage of the lane line in the conventional method can be solved.

The invention provides an image enhancement and abstraction method based on anisotropic filtering. The method comprises the following steps: calculating the smooth continuous characteristic-preserved edge tangential flow field of an input image by use of a structure tensor smoothing technology; performing adaptive smoothing on the image along the direction of the edge tangential flow; performing unsharp image enhancement based on a Gaussian difference mask along the gradient direction; and performing improved bilateral filtering based on characteristic flow or improved Kuwahara filtering based on characteristic flow to obtain a characteristic-enhanced abstract image. The method provided by the invention can obviously improve the generation quality of the existing image abstraction method; and the obtained abstract image has obviously-enhanced boundary characteristics, clear image quality and distinct layers.

The invention discloses an earthquake image structure guiding noise reduction method based on regularization mixed norm filtering. The earthquake image structure guiding noise reduction method includes the following steps that a gradient structure tensor is solved for an input three-dimensional earthquake image; regularization mixed norm filtering is conducted on the gradient structure tensor; a diffusion tensor is designed according to the eigenvalue and eigenvector of the filtered gradient structure tensor; continuity factors are calculated, the continuity factors at the position of a boundary fault feather edge and the like are close to zero, and the maintain performance of the structure is achieved; a sobel operator serves as a derivation operator so that divergence can be calculated. By means of the earthquake image structure guiding noise reduction method based on regularization mixed norm filtering, the textured edge information of earthquake-related data can be reserved, Gaussian noise, ultra Gaussian noise and sub Gaussian noise can be effectively suppressed, and therefore an efficient noise reduction method is achieved.

The invention discloses an optical remote sensing image region-of-interest detection method which is applied to high-resolution optical remote sensing image region-of-interest detection. A super-pixel-level image is processed, super pixel-to-pixel saliency mapping is performed to get a pixel-level saliency map, and thus, a region of interest is quickly located from coarse-scale description to fine-scale precise description. The method comprises the following steps: firstly, down-sampling an original image, performing super pixel segmentation, and converting the high-resolution image into a coarse-scale image based on the operations; secondly, getting a super-pixel-level texture feature map and a super-pixel-level color feature map through a structure tensor and a color space background suppression technology based on the generated coarse-scale image; thirdly, getting a super-pixel-level saliency map under the original resolution scale through feature map fusion and up-sampling interpolation; and finally, getting a pixel-level saliency map through super pixel-to-pixel saliency mapping, thus completing pixel-level precise description of the region of interest.

The invention discloses an adaptive denoising method for an ultrasonic image. The adaptive denoising method comprises the following steps: (1) obtaining the ultrasonic image; (2) performing Gaussian-Laplacian pyramid decomposition on the ultrasonic image and obtaining Gaussian layers and Laplacian layers at different scales; (3) calculating a structure tensor and a diffusion tensor of the Gaussian layer at each scale and performing anisotropic diffusion filtering treatment on the Gaussian layer at the scale; (4) according to a characteristic value of the structure tensor of the Gaussian layer, designing a grey mapping curve, and performing grey mapping on the Gaussian layer at the scale according to the grey mapping curve; (5) repeating the steps (3) and (4) for multiple times and performing the same treatment on the Gaussian layer and the Laplacian layer at each scale; (6) performing reverse reconstruction on the treated Gaussian layer and Laplacian layer and obtaining a denoised ultrasonic image; (7) outputting the denoised ultrasonic image. The adaptive denoising method can inhibit edge enhancement and spots of the image and is simple in algorithm and strong in adaptability.

The invention relates to an image denoising method integrated with a structure tensor and non-local total variation, and belongs to the technical field of image processing. The method comprises the following steps: 1, performing N*N portioning on an image f with noise, wherein the value of N can be 3, 5, 7...; 2, performing denoising modeling on the image to establish an object function integrated with the structure tensor and the non-local total variation; and 3, obtaining a denoised image by solving the object function by use of a split Bregman algorithm. The method provided by the invention has the following advantages: texture features and geometric structure features of the image are maintained while noise is effectively eliminated, the algorithm is rapid, and at the same time, a step effect can be reduced.

The invention discloses a light field foreground segmentation method and device based on K-means clustering. The method includes the steps: extracting a refocusing image, a polar line plane image anda full clear image from a light field image to be processed; processing the polar line plane image by a structure tensor method to acquire polar line plane depth information; processing the refocusingimage by a discrete cosine response method to acquire refocusing information; acquiring area color features, area geometric features, area corresponding point features and area refocusing features according to each of a plurality of areas divided from the full clear image by a super-pixel segmentation technology; calculating similarity of the areas by K-means clustering; marking a foreground anda background by a graph cut algorithm based on the similarity to acquire foreground segmentation results of the light field image. The foreground segmentation results processed by the method are moreaccurate than those in the prior art.

The invention discloses a stratum structure self-adaption median filtering method. The method comprises the steps of: 1, calculating a gradient structure tensor; 2, calculating a stratum lateral incontinuity measurement; 3, structuring a structure self-adaption median filtering device; and 4, performing structure self-adaption median filtering and processing. By utilizing the method, earthquake random noise and part coherent noise can be effectively attenuated, and stratum edge and detailed structure characteristics such as effective signals, geologic faults and amplitude abnormalities can be maintained in a filtering process; the algorithm of the technical scheme is easy to realize and has good operability; and moreover, a selection problem of a median filtering processing window size is avoided, the method meets the requirements of the median filtering processing of complicated signals, for data areas with good event extension, the filtering performance of the median filtering device is improved through adaptively controlling the direction characteristics of a filtering window, and for areas with geologic faults and angular unconformity, the maintaining performance of the median filtering device to stratum edge structures such as the geologic faults is improved through adaptively adjusting the size of the filtering window.

The invention discloses a video denoising method based on improved self-adaptive bilateral filter. The video denoising method includes that: 1) obtaining frames at two ends and the center of continuous N frames, and calculating a characteristic value of structure tensor of an image; 2) performing statistics on proportion of edge areas in the image and gradient distribution of the edges; 3) calculating gray average of pixels according to the direction of pixel structure tensor; 4) dividing an edge area and a non-edge area; 5) arranged gray variance in the edge area according to gradient size of the edges; 6) arranged large gray variance in the non-edge area; and 7) calculating the bilateral filter according to the self-adaptive gray average and the gray variance. The video denoising method based on the improved self-adaptive bilateral filter uses the front frames of a video as reference for edge statistics, reduces calculated amount, and enhances stability of the filter with the help of the characteristic that edge areas between adjacent frames are overlapped.

The invention relates to an image fusion and super-resolution achievement method based on variation and fractional order differential, and belongs to the field of image processing and information fusion. On the basis of image fusion and super-resolution achievement, a low-resolution source image to be fused is regarded as a multi-channel image, unit value representation of gradient characteristics of the multi-channel image is obtained through construction of structure tensor of the low-resolution source image, and an image fusion and super-resolution achievement model is established according to the same or similar gradient characteristics between the low-resolution fusion image and the multi-channel image; in the model, the fractional order differential and fractional order total-variation minimization achievement method is introduced to achieve noise suppression, image edge information is enhanced through diffusion of two-way filtering wave diffusion, and generation of false information is suppressed. The method overcomes the defect that with a traditional method, fusion and super-resolution achievement can not be performed at the same time and has good application prospects in the fields of target imaging, safety monitoring and the like.

Provided is a single-image super-resolution reconstruction method based on edge difference constraint. The method includes following three steps: step 1, extracting a texture principal direction characteristic of a training image through a Gabor filter, and performing a principal component analysis dictionary training to obtain a training dictionary; step 2, constructing a reconstruction model by employing the dictionary, and obtaining an initial reconstruction high-resolution image with a good edge structure through iterative threshold shrinkage; and step 3, describing an operator, a spatial distance, a pixel intensity, and edge orientation information by employing a histogram of oriented gradients between image blocks, establishing a non-local structure tensor optimization model, further optimizing and processing the initial reconstruction high-resolution image, and obtaining a final reconstruction high-resolution image with a substantial edge structure and abundant detail information. According to the method, by considering the difference between the initial reconstruction high-resolution image and an original clear image, the post-processing optimization method is further proposed, and the detail information of image edges and textures is abundant.

The invention discloses a weight window self-adaptation non-local mean image denoising method. According to the weight window self-adaptation non-local mean image denoising method, the sizes of weight windows can be controlled in a self-adaptation mode according to image local structure characteristics, noise is suppressed while an edge structure is protected, and therefore the image quality is remarkably improved. The method includes the following steps that first, a frame of noise image is initialized and read in; second, a structure tensor matrix is built; third, according to the built structure tensor matrix, edge structure indicators are built, and the characteristics of the area where pixel dots are located are positioned; fourth, the areas of the image are classified through the edge structure indicators; fifth, according to the type of the area to which each pixel dot belongs, the size of the adjacent area of each pixel dot is determined; sixth, according to the determined size of the adjacent area of each pixel dot, a similarity metric function between the adjacent areas is built; seventh, S dots with highest similarity are screened; eighth, a denoising model is built, and a denoised image is acquired.

The present invention relates to the technical field of seismic data processing, and in particular to a seismic data edge preserving smoothing method based on structural features, which performs targeted sub-region de-noising treatment based on the structural information of seismic data and the similarity information of the pre-filtering process. In addition to the random noise removal, the method effectively protects the boundary information of seismic data. The edge preserving smoothing method includes steps of smoothing the seismic data, decomposing the matrix to obtain the structure tensor of the seismic data, and completing the structural feature analysis of the seismic data; using the conventional de-noising method to filter the seismic data and obtaining the local similarity coefficient; building a guide filtering equation; and performing multiple de-noising iterations and other steps.

The invention discloses a calculating method of a visual salience drawing. The method comprises the steps of adopting an Itti model to generate a color salience drawing and a brightness salience drawing of a drawing; effectively extracting an edge salience drawing and an angular point salience drawing by using structure tensors; generating an outline salience drawing through linear combination; and finally carrying out a normalized and linear combination through characteristic salience drawing to obtain a final salience drawing. According to the calculating method provided by the invention, the visual salience drawing obtained is clearer, a significant target is more accurate to locate, and the efficiency calculating aspect can be improved.

The invention discloses an infrared weak small target detection method based on tensor robust principal component analysis, and relates to the field of infrared image processing and target detection.The method comprises the steps of 1 traversing an original image to construct a third-order tensor; 2 calculating a second-order structure tensor of the original image, and constructing a structure weight tensor; 3 using the tensor robust principal component analysis for constructing an objective function, inputting a third-order tensor and a structure weight tensor into the objective function, and using an ADMM for solving the objective function to obtain a background tensor and an objective tensor; 4 reconstructing a background image and a target image according to the background tensor andthe target tensor; 5 segmenting the target image and outputting a target detection result. According to the method, the problem that the target detection accuracy is low due to the fact that the nuclear norm and the local structure weight adopted in an existing method easily cause local optimal solution and detection target distortion is solved, and the effects of improving the target detection and background inhibition capability and enhancing the target shape keeping capability are achieved.

The invention discloses a three-dimensional seismic data image denoising method based on trilateral structure guide smoothing. A frame based on traditional structure guide smoothing is included, the construction of structure tensor is improved, the trilateral smoothing with a good side protection effect is combined, the performance of original Gaussian kernel anisotropic diffusion smoothing is greatly improved, the denoising effect is good, less structure information is lost, a great effect is achieved in the side protection and denoising of a seism image, the seism image can store more structure information in the denoising process, and the explanation and the subsequent processing of the seism image are facilitated.

The present invention discloses a face image illumination compensation method. The method comprises the following steps that: step 1, classical retinex illumination compensation is performed on an original image; step 2, edge detection is performed, a pseudo edge is judged by using an equation (13), and a low-illuminance region C (x, y) corresponding to the pseudo edge is marked by using an equation (17); step 3, structural tensors and eigenvalues mu1 and mu2 corresponding to the structural tensors are obtained according to equations (14) and (16); step 4, values are re-assigned to mu1 and mu2, the values obtained in the step 2 and step 3 are introduced into an equation (20); and step 5, the environment function of the Retinex algorithm is improved, illumination processing is performed on the original image, and an illumination-compensated image is obtained by using an equation (5). With the method of the invention adopted, the inadequacies of traditional Retinex and PCNN in eliminating shadows and causing fogging phenomena in illumination compensation can be eliminated, shadows in the image can be diluted to a certain extent, the shadow edges of the image can be eliminated, detail information can be presented, a face recognition rate can be improved, and a face misjudgment rate can be decreased.

The invention discloses a method for detecting the contour of a carbonate rock fault-karst body, comprising: constructing a gradient structure tensor based on a three-dimensional seismic data body and calculating the eigenvalues ​​of the gradient structure tensor; determining the gradient structure according to the characteristics of the actual fault-karst body A smoothing parameter of the eigenvalue of the tensor, and using the smoothing parameter to smooth the eigenvalue of the gradient structure tensor; and using the horizontal well drilling time curve of the area related to the carbonate rock fault-karst body to calibrate the feature value, so as to obtain the outline of carbonate fault-karst bodies.