79 results about "Curvelet" patented technology

The invention discloses a short-term power load forecasting method in the technical field of power load forecasting. The present invention constructs a sample set through the load data of the data collection and monitoring control system, and denoises the sample set through curvelet transform to obtain a denoised sample set; divides the denoised sample set into a test set and a training set; Use the training set and learning machine to generate multiple training models, and then use the bagging algorithm to obtain the final prediction model; finally use the final prediction model and test set to predict the load. The invention not only solves the problems of small amount of sample data, large deviation and uncertainty, but also has stronger generalization ability than a single learning machine, can effectively integrate multiple models, and makes the prediction process more rapid and accurate.

The invention discloses a vehicle type recognition method based on vehicle face features. The method comprises the following steps of: (1) collecting vehicle images through a monitoring camera, preprocessing the images and detecting and segmenting vehicle face images capable of characterizing vehicle types, wherein the step (1) comprises the following specific procedures of: (1-1) collecting the vehicle images of various types of vehicles by using the monitoring camera, (1-2) preprocessing the images and enhancing the quality of the images through homomorphic filtering, and (1-3) segmenting vehicle face regions capable of characterizing vehicle type features by adopting vehicle face region detection and segmentation based on license plate position information; (2) carrying out Curvelet wavelet transform on the vehicle images so as to extract a vehicle face feature matrix capable of characterizing the vehicle type features; and (3) classifying the Curvelet wavelet feature vectors of the extracted vehicle face images by using a support vector machine classifier so as to recognize the vehicle type. The method provided by the invention can be used for providing more accurate vehicle and vehicle type information for traffic monitoring and is very important for traffic safety and the real-time extraction of traffic information.

The invention provides a medical ultrasonic fundamental wave and harmonic wave image fusion method, which comprises the following steps: firstly, performing Curvelet decomposition of an ultrasonic fundamental wave and harmonic wave image to obtain a Curvelet coefficient; secondly, performing fusion treatment of the Curvelet coefficient, and obtaining a fusion Curvelet coefficient by using a weighted average method for a low-frequency part and an absolute value maximum selection method for a high-frequency part; and thirdly, reconstructing a fusion result image by Curvelet inverse transformation according to the Curvelet coefficient obtained by the fusion treatment. In the invention, according to the characteristics of the ultrasonic fundamental wave and harmonic wave image, the fusion of the ultrasonic fundamental wave and harmonic wave image is performed by using a Curvelet method to obtain an image with clear organizational boundaries and interior, the problems of blurred edges, difficult organization positioning and the like of common imaging harmonic wave imaging are solved, and the method can be widely used in the processing of medical ultrasonic images.

The invention discloses a no-reference objective three-dimensional image quality evaluation method based on binocular visual perception. The method comprises the steps of constructing a converging one-eyed image of a distorted three-dimensional image by using an energy gain control model, and constructing left and right disparity images and indefinite left and right images by using left and right viewpoint images; then extracting a curvelet domain feature from the converging one-eyed image, and separately extracting a generalized Gaussian fitting parameter feature and a lognormal distribution fitting parameter feature from the left and right disparity images and the indefinite left and right images, wherein the three features are used as three-dimensional image feature information; and finally, constructing a relation between three-dimensional image features and average subjective scoring differences through support vector regression to obtain an objective quality evaluation predicted value of the distorted three-dimensional image. The method has the advantages that the acquired feature vector of the distorted three-dimensional image has strong stability and can reflect the quality change condition of the distorted three-dimensional image, the objective evaluation has good consistency with subjective perception of human eyes, and the correlation between the objective evaluation result and the subjective perception is improved.

The invention discloses a false fingerprint detection algorithm based on curvelet texture analysis and SVM-KNN classification. The false fingerprint detection algorithm based on the curvelet texture analysis and the SVM-KNN classification comprises the following steps of curvelet transformation; curvelet reconstruction; curvelet coefficient feature extraction, extraction of coefficient energy, entropy and the like, and coefficient first-order statistical magnitude extraction; textural feature extraction, and extraction of a first-order statistical magnitude, a gray-level co-occurrence matrix, MRF features and the like; training of classifiers; performance evaluation of the classifiers; false fingerprint detection, wherein testing is carried out on samples by utilizing the trained classifiers. The false fingerprint detection algorithm based on the curvelet texture analysis and the SVM-KNN classification is suitable for various fingerprint acquisition instruments of 500 dpi resolution, analysis is respectively carried out on image high-frequency noise information and texture information after denoising, and the noise and texture differences between true fingerprints and false fingerprints are quantified.

The invention relates to the seismic exploration field, especially to a random noise suppression method and apparatus for seismic data. The method comprises: complementary ensemble empirical mode decomposition (CEEMD) is carried out on seismic data to obtain an intrinsic mode function (IMF) component sequence, and according to an interrelation between seismic data and IMF components, IMF components including random noises and IMF components not including random noises are determined; on the basis of differences of the random noises included by the IMF components including random noises, different threshold values are selected for the IMF components including random noises to carry out optimal curvelet iteration threshold de-noising processing, thereby obtaining the processed IMF components; and then seismic data after noise removing are obtained by reconstruction. According to the method provided by the embodiment of the application, on the basis of combination of the CEEMD and the optimal curvelet iteration threshold method, defects of effective signal loss due to an EMD method and de-noising method and single threshold selection because of a curvelet threshold de-noising method can be overcome. The effective signal can be kept well while random noises are suppressed, so that the signal to noise ratio of seismic data is improved.

The invention discloses an extraction method for multi-scale multi-direction textural features of froth images. First, froth grayscale images are subjected to curvelet transformation; then, curvelet subgraphs with different scales in different directions are processed respectively, and multi-scale multi-direction textural representation information is extracted to form feature vectors of the froth images; the froth images with different working conditions can be distinguished according to obtained textural features. The extraction method for the multi-scale multi-direction textural features of the froth images has good pattern separability with respect to identification of the froth images, and is easy to implement.

The invention discloses a fixed wing unmanned aerial vehicle touring image accurately-splicing method for a power transmission line. The accurately-splicing method includes the following steps: (1) conducting data preparation and sample training; (2) conducting feature extraction on the power transmission line based on textural features and straight line features to obtain a power transmission line distributing area and power transmission line extraction data, screening the power transmission line extraction data through the power transmission line distributing area, and obtaining power transmission line extracting results through power transmission line double-edge features; (3) conducting SIFT image registration and fusing based on power transmission line customization on the power transmission line extracting results. According to the fixed wing unmanned aerial vehicle touring image accurately-splicing method, curvelet transformation is adopted for achieving extraction on power transmission line texture information and detection on power transmission line distribution, and the power transmission line double-edge features in high-resolution unmanned aerial vehicle images are further adopted for obtaining the reliable and accurate power transmission line extracting results.

The invention relates to a seismic data denoising method based on contourlet transformation. The seismic data denoising method is characterized by comprising the following steps that firstly, seismic data are read, and Laplacian pyramid decomposing is conducted; secondly, high-frequency sub-bands of all dimensions are obtained through the Laplacian pyramid decomposing and are input into directional filter banks, and components of the high-frequency sub-bands of all dimensions in all directions are obtained; thirdly, a noise model is selected according to features of noise in seismic signals, pyramidal direction filter bank decomposing is conducted on the model, a threshold value is obtained, and the threshold value is used for conducting filtering on the components of the high-frequency sub-bands of all dimensions in all directions; fourthly, pyramidal direction filter bank inverse transformation is conducted on the components of the filtered high-frequency sub-bands of all dimensions in all directions, and denoised seismic signals are obtained. According to the seismic data denoising method based on contourlet transformation, contourlet transformation is applied to seismic data processing, the defect that useful signals are damaged while denoising is conducted through methods of wavelet transformation, curvelet transformation and the like is overcome, the fidelity of the seismic data is improved, noise is suppressed, and effective information is extracted.

The invention discloses an image sparse representation method based on a Curvelet redundant dictionary, mainly aiming to solve the problems that in the existing method, the redundant dictionary has large scale, the calculation complexity is high, and sparse representation can not be effectively carried out on the rich border outline details in the image. The invention is realized through the following steps: (1) selecting the tight frame of Curvelet as an atomic model; (2) determining the numeric areas of the scale parameter j, direction parameter theta and displacement parameter k in the frame, carrying out discretization on each parameter to form the Curvelet redundant dictionary; and (3) blocking each input image, carrying out sparse decomposition on each sub-image by utilizing an orthogonal matching pursuit (OMP) algorithm sparse decomposition to solve sparse coefficient vectors, combining all the sparse coefficient vectors to obtain the sparse matrix, and multiplying the sparse matrix by the Curvelet redundant dictionary to obtain the sparse representation results of the input image. Compared with the prior art, the invention has the advantages of low calculation complexity, high quality of sparse representation image, especially can better capture the singularity of curves in the image, and can be applied to the fields of image processing and computer vision.

Method for adapting a template to a target data set. The template may be used to remove noise from, or interpret noise in, the target data set. The target data set is transformed (550) using a selected complex-valued, directional, multi-resolution transform (‘CDMT’) satisfying the Hubert transform property at least approximately. An initial template is selected, and it is transformed (551) using the same CDMT. Then the transformed template is adapted (560) to the transformed target data by adjusting the template's expansion coefficients within allowed ranges of adjustment so as to better match the expansion coefficients of the target data set. Multiple templates may be simultaneously adapted to better fit the noise or other component of the data that it may be desired to represent by template.

The invention discloses a polarized image fusion method based on discrete continuous curvelet, and belongs to the field of image processing. The method comprises the following step: firstly, performing discrete continuous curvelet to polarization strength and polarization ratio images to obtain low-frequency sub-band coefficients and each direction sub-band coefficient; then selectively fusing low-frequency sub-band coefficient for the low-frequency sub-band coefficient by using weighted average criterion, selectively fusing each direction sub-band coefficient for each direction sub-band coefficient by using area energy maximal criterion, and finally performing inverse discrete continuous curvelet to obtain a final fusion image. The discrete continuous curvelet is realized fast by adopting the Wrapping-based method, and the redundant information of the conversion result is low. Experimental results prove that the algorithm disclosed by the invention is very effective; furthermore, the edge and space grain information of the fused image are clear, the computing time of the algorithm is short, and the image information can be displayed well in real time.

The invention discloses a multi-sensor video fusion method based on space-time conspicuousness detection. The method comprises the steps of respectively inputting two registered videos; utilizing three dimensional uniform discrete curvelet (3D-UDCT) to decompose so as to obtain sub-band coefficients; dividing a video area into three different areas; combining different areas according to different fusion strategies so as obtain high-pass direction sub-band coefficients of a fused video; conducting weighted average on low-pass sub-band coefficients so as to obtain low-pass sub-band coefficients of the fused video; and conducting 3D-UDCT inverse transformation so as to obtain the fused video. The method overcomes the defects of information extraction limited by space and space-time uniformity in the prior art, is capable of better extracting conspicuous space-time characteristic information input in video images so as to enable the video to be better in space-time uniformity and stability, is good in noise robustness, and can be used in video image fusion under a static background.

The invention discloses a synthetic aperture radar (SAR) image target identification method based on curvelet domain bilateral two-dimension principal component analysis. The method specifically comprises the following steps of: inputting images of a training sample and a test sample, and normalizing the sample images; performing curvelet transformation on the normalized samples, and extracting low-frequency sub-band coefficients of each sample which is transformed; acquiring left and right projection matrixes of characteristics according to the obtained low-frequency sub-band coefficients of the training sample; acquiring characteristic values of the training sample and the test sample by using the left and right projection matrixes which are obtained; and classifying the characteristics of the test sample by using a nearest neighbor classification method, and thus obtaining a final identification result. Compared with the prior art, the method has the advantages that the dimensionality of the characteristics is effectively reduced, high correct identification rate can be obtained, an implementation method is simple, and identification time is effectively shortened.

The invention discloses a non-homogeneous curvelet three-dimensional earthquake data reconstruction method based on a linear Bregman algorithm. The method is characterized by comprising steps of firstly successively extracting time slices of non-homogeneous three-dimensional earthquake data in two space directions; based on multi-dimension multi-direction two-dimensional curvelet positive conversion, introducing two-dimensional space non-homogeneous rapid Fourier transformation; establishing non-homogeneous curvelet inverse conversion operators between earthquake missing data under homogeneous curvelet coefficients and space non-homogeneous sampling; using the linear Bregman algorithm to carry out resolving; by selecting proper threshold value factors and dynamic step lengths, and adopting soft threshold value operators, carry out precise back calculation to obtain homogeneous curvelet coefficients of irregular earthquake data under non-homogeneous sampling; and finally, carrying out normal curvelet inverse transformation, thereby forming a new reconstruction method. According to the invention, the resolution ratio and the signal to noise ratio of reconstructed signals are greatly improved; and the method has important value in aspects of guiding acquisition of non-homogenous earthquake data and reconstruction of missing roads in complex regions.

The present invention provides a system and a method for eliminating the random noise in seismic signals, which comprises a curvelet transformation unit, a threshold optimization unit, a threshold de-noising unit, a filter design unit, a spectral whitening processing unit, and a curvelet inverse-transformation unit. The curvelet transformation unit is used for reading original seismic signal data, and obtaining first frequency sub-bands and second frequency sub-bands in various scales and all directions within a curvelet domain through decomposing the data based on the wavelet transformation process. The threshold optimization unit is used for determining the threshold values of curvelet coefficients in various scales and all directions, and obtaining an optimized threshold based on the genetic algorithm, wherein the optimized threshold is set in such a manner that the risk assessment function of the generalized cross validation criteria has a minimum value. The threshold de-noising unit is used for conducting the threshold de-noising treatment on the first and second frequency sub-bands obtained through the decomposing process. The filter design unit is used for designing different self-adaptive spectral whitening filters. The spectral whitening processing unit is used for conducting the spectral whitening treatment on effective signals of the first frequency sub-bands and the second frequency sub-bands in various scales and all directions. The curvelet inverse-transformation unit is used for reconstructing the signals through the curvelet inverse-transformation process to obtain de-noised seismic signal data of higher resolution.

The invention relates to a morphological component analysis (MCA)-based synthetic aperture radar (SAR) image noise suppression method. The method is technically characterized in that: an MCA method is applied to the suppression of multiplicative noises in an SAR image; and because the image generally comprises the two morphological components of edge / contour and texture, in the method, the dictionary assembly of curvelet and local discrete cosine transform (LDCT) is utilized under the uniform frame of MCA according to the good edge / contour representation characteristics of the curvelet and the high partial texture expression capability of the LDCT to fulfill the aim of suppressing the noises in the SAR image. Therefore, for a noise image, effective components of the image can be separated and the noises are kept in the remnant image by the method to fulfill the aim of suppressing the noises. Experimental results on both simulated and real SAR images show that the method has the advantage of relatively better noise suppression effect compared with the conventional wavelet-based methods and curvelet-based methods.

The present invention relates to a multi-component seismic data Corssline direction wavefield reconstruction method based on Shearlet transform, which attributes the seismic data reconstruction to l1 regularization based on Shearlet sparse constraints; utilizes the simultaneous observation of pressure and acceleration in multi-component measurement, and the gradient is The reconstructed seismic data provides additional information, which can reconstruct the data more accurately; the P and acceleration wavefields are reconstructed in an alternate iterative manner. The fundamental difference between this method and other methods is that it expands to multi-component seismic data interpolation on the basis of single-component seismic data interpolation. Since the addition of velocity components increases constraints, better reconstruction results can be achieved. It solves the problem that the existing single-component seismic Crossline direction seismic wavefield reconstruction method has low precision, cannot effectively reconstruct small underground structures, and is difficult to reconstruct. Solved the problem that the reconstruction effect of Curvelet with large curvature is not good. The cost is reduced and the accuracy is improved.

The invention relates to a noise image enhancing method based on non-linear Curvelet diffusion, which is characterized in that the blurry edge and positioning problems are avoided occurring in isotropic diffusion by utilizing the heat diffusion in the nature in the isotropic diffusion process based on a PDE (Partial Differential Equation) method. By using the method, the noise removing and the edge maintaining are synchronously realized, but the method is very sensitive to the noise. Aiming at the noise image, the invention provides an enhancing method combining Curvelet (Mirror-extended Curvelet, ME-Curvelet) conversion with non-linear diffusion. By using the method, the edge character of image and the contrast are enhanced while the noise therein is restrained, the fake Gibbs effect can be efficiently reduced, and the image quality is further increased.

The invention discloses an image block effect removing method on the basis of curvelet transformation, which comprises the following steps of: step 1, carrying out curvelet transformation on a target image to obtain scale levels and a curvelet coefficient of each level; step 2, finding a layer which is furthest influenced by a block effect and processing the layer; and step 3, recovering the image. According to the invention, the curvelet transformation is utilized to carry out layered processing on the image; the recovered image is reconstructed by a novel coefficient matrix; and compared with a conventional block effect removing method, the image block effect removing method disclosed by the invention has a better recovery effect on the aspects of both subjective and objective evaluation.

The invention discloses a rapid fuzzy matching method used in high-speed railway catenary rod insulator adverse-state detection. Aiming at problems of electrified railway catenary insulator ceramic bottle breakage fault detections, insulator characteristic quantities are obtained by anisotropic directional filtering upon comprehensive vehicle inspection field images by using second-generation curvelet. For solving a problem that the ceramic bottle characteristic coefficient distributions of different insulators are irregular, a method for balancing curvelet coefficients by utilizing directional morphological closing operation is provided. For solving a problem of intelligent identification, a method of dimension reduction upon curvelet coefficient matrix by utilizing an energy band method is provided. With the method provided by the invention, insulator position information can be obtained, and insulator fault can be determined. As a result of experiments, with the method, global analysis can be directly carried out upon images shot in field. Relatively high accuracies are achieved in insulator positioning in the images, and in insulator fault determinations. Therefore, a novel means is provided for electrified railway insulation reliability detections.

The invention discloses a curvelet redundant dictionary based immune optimization image reconstruction method, which solves the problem that the present 10-norm reconstruction technology acquires a reconstructed image with a poor visual effect, and is realized through the following steps: (1) clustering observation vectors; (2) initializing populations; (3) para-immunity optimizing; (4) reconstructing an initial image; (5) filtering and projecting onto convex sets; (6) judging whether or not the iteration number achieves the maximum value; (7) updating the sparsity; (8) updating the populations; (9) immune optimizing an image block; and (10) reconstructing the image. According to the invention, similar observation vectors are solved to obtain a common set of curvelet ground atoms by using the immune clone optimization technology, and then each observation vector is solved to obtain a set of curvelet ground atoms through combining the filtering and the projecting onto the convex sets. The curvelet redundant dictionary based immune optimization image reconstruction method eliminates the block effect in the reconstructed image, and obtains a reconstruction image with better visual effect.

The invention provides a self-adaption curvelet threshold value earthquake denoising method based on local variance analysis. According to the method, on the basis that deep analysis is conducted on the characteristics of a curvelet coefficient on different scales and at different angles, a self-adaption threshold value technology is provided so that effective separation of a signal and noise can be achieved at any angle, protection conducted on non-noise components is enhanced with the help of local variance information of a curvelet area, and therefore damage to the effective signal especially the weak signal can be reduced. Practices on synthetic seismogram and land seismic data prove that the actual denoising effect of a threshold value iteration method is not ideal, the denoising effect of the method is more real and effective, and the method has the unique advantages in the aspects of noise attenuation and detail protection.

The invention discloses an image fusion method based on multi-morphological sparse representation. The method includes the steps of 1, extracting edge components of an infrared image by means of frequency-tuned salient region detection, and subjecting the edge components of the infrared image to spare representation by means of Curvelet basis; 2, forming a multi-morphological sparse basis by means of DWT (discrete wavelet transform) and DCT (discrete cosine transformer), and subjecting a visible light image to spare representation by means of the multi-morphological sparse basis so as to obtain spare coefficients of image smooth components and texture components; 3, fusing the edge components of the infrared image to the spare coefficients of the smooth components and texture components of the visible light image; 4, subjecting the fused sparse coefficients to inverse transform so as to obtain a final fused image. The method has the advantages that image data quantity is decreased, the edge information and texture information of the image is fully retained, and quality of the fused image is greatly improved.

The invention discloses an extraction method for an in-phase axis slope in seismic data. The extraction method comprises the following steps that step one, the seismic data are acquired; step two, discrete curvelet transformation is performed based on the seismic data so that a series of curvelet transformation coefficients are obtained; step three, inverse transformation is performed based on the main curvelet transformation coefficient of a series of the curvelet transformation coefficients so that the curvelet carrying directional information is obtained; and step four, the in-phase axis slope is extracted according to the curvelet carrying the directional information. Another embodiment of the invention provides an application of the extraction method in seismic data processing and imaging. According to the extraction method, the accurate slope can be obtained so as to have great effect in seismic data processing and imaging.

The invention discloses a high-resolution underground structure amplitude-preserving imaging method, and the method comprises the steps: obtaining an initial imaging result through reverse time migration, carrying out the Born forward modeling on the basis of the initial imaging result to obtain simulation seismic data, and carrying out the reverse time migration on the simulation data to obtain asecondary migration imaging result; then carrying out curvelet transformation on the two imaging results, and carrying out point-by-point estimation in a curvelet domain, wherein Wiener solutions matched with the two groups of curvelet coefficients serve as solutions of a matched filter; and finally, acting the estimated matching filter on the initial imaging result to obtain a high-resolution amplitude-preserving imaging result. The method aims at solving the problems that imaging results are fuzzy and amplitude is unbalanced due to the fact that a reverse time migration imaging operator isan accompanying operator of a forward operator. And migration imaging is performed again on the initial migration imaging result, a matched filter of two migration imaging results in a transform domain is searched by utilizing curvelet transform to serve as approximation of an inverse Hessian operator, and the matched filter acts on the initial imaging result so as to achieve the effect of improving the imaging quality.

The invention proposes a curvelet domain enhancement method for a low-illumination-level power equipment image, and solves a problem that the degradation of a low-illumination-level power equipment image is severe. The method comprises the steps: converting the low-illumination-level power equipment image into an HIS space, decomposing a brightness parameter through curvelet transformation, and obtaining subband components at different scales and in different direction, so as to construct a human eye vision model; carrying out the nonlinear enhancement of a high-frequency component through brightness shading and brightness-contrast shading characteristics of the model, and carrying out the nonlinear stretching of a low-frequency component; finally reconstructing the brightness parameter through the curvelet inverse transformation, converting the image into an original color space through combining with chromaticity and saturability components of the original image, and obtaining an enhanced low-illumination-level power equipment image. The method can effectively improve the contrast and brightness of the low-illumination-level power equipment image, maintains the detail informationof the image, and inhibits the image noises.

The invention relates to a method and a system for image fusion based on Curvelet transformation. The method comprises the following steps: interpolating an original hyper-spectral image to get a new hyper-spectral image; carrying out HIS transformation to get components I, S and H, and stretching the gray of an SAR image to get an SAR* image; carrying out Curvelet transformation, and carrying out fusion according to fusion rules to get a new Curvelet transformation coefficient; and finally, carrying out inverse Curvelet transformation to get a component I*, and carrying out inverse HIS transformation on the component I* and the components S and H to get a fused image. The invention further relates to a system which comprises an image interpolation module, an image processing module, a Curvelet transformation module, and an image fusion module. The spectral distortion of fused images is improved, and the spatial resolution is increased greatly.

The invention belongs to the technical field of the data de-noising, and discloses a curvelet domain statistics self-adaptive threshold ground penetrating radar data de-noising method and system. Themethod comprises the following steps: importing a massive complex domain threshold function algorithm, analyzing a change rule of the traditional threshold function curvelet conversion de-noising effect along a threshold function control coefficient so as to be used for the subsequent curvelet domain statistics self-adaptive threshold contrast; performing correlation superposition on the curveletconversion coefficient on the scale and direction by utilizing high-order statistics theory, and statistically and self-adaptively determining the distribution scale and rotational direction of an effective signal on the curvelet conversion coefficient through the correlation; determining a noise removing component threshold range, constructing a statistics self-adaptive threshold function curvelet conversion de-noising algorithm. Compared with the prior art, the processing result on the synthetic ground penetrating radar data containing random noise and related noise and the actually-measuredground penetrating radar data has guidance significance on the precise inference interpretation on the complex ground penetrating radar data.

The invention provides a CS image denoising reconstruction method based on hyperspectral total variation. The CS image denoising reconstruction method comprises the following steps: initializing a reconstructed image, an iterative index value and a noisy observation value; iteratively updating the obtained reconstructed image by using the noisy observation value to obtain an estimated value; respectively inputting the estimated values into a CS reconstruction model based on the l1-norm and the HTV to obtain an intermediate reconstruction image; performing sparse representation on the intermediate reconstructed image by using Starlet transform to obtain a Starlet coefficient; performing denoising filtering on the Starlet coefficient by using the new threshold operator and the improved BayeShrink threshold to obtain a curvelet coefficient; performing Starlet inverse transformation on the curvelet coefficient to obtain a reconstructed image; and judging whether an iteration stopping condition is met or not, and carrying out loop iteration. According to the method, while most noise information in the high-noise image is removed, details, textures and other feature information in the image can be effectively protected, the method is easy to implement and high in robustness, and the denoising reconstruction problem of the high-noise image is effectively solved.