34 results about "Mellin transform" patented technology

The present invention provides a method and system for determining near-duplicate images. The method and system includes performing a Fourier-Mellin transform on each of a plurality of images. For each image of the plurality of images, the method and system includes generating a signature based on the Fourier-Mellin transform. The method and system includes comparing the signature of at least one of the images to at least one of the signatures of the other plurality of images and determining any near duplicate images based on the comparing of the signatures.

The embodiment of the invention discloses an image tampering detection method and device. The method comprises the following steps: segmenting a to-be-detected image into multiple sub-images with overlapping regions by using a sliding window; extracting feature vectors of various sub-images by using a discrete analytic Fourier-Merlin transform method; performing multiple projections on various feature vectors by using a local sensitivity Hash algorithm to respectively multiple projection values with the same number of each sub-image; judging whether a target sub-image pair meeting a preset projection condition is existent; regarding that the to-be-detected image is not tampered if the target sub-image pair meeting the preset projection condition is nonexistent; computing a filter referencevalue of each target sub-image in the target sub-image pair by using a Weber local description sub-method and a variance function if the target sub-image pair meeting the preset projection conditionis existent, and judging whether each filter reference value is greater than a threshold; regarding that the to-be-detected image is tampered if the filter reference value is greater than the threshold; or regarding that the to-be-detected image is not tampered if the filter reference value is not greater than the threshold. By using the technical scheme provided by the application, the accuracy of image tampering detection is improved, the computing complexity is reduced, and the recall ratio is improved.

The invention relates to a Gamma-Gamma distribution parameter estimation method based on a double logarithmic cumulant expectation, and belongs to the technical field of atmospheric turbulence channel parameter estimation of free space optical communication systems. According to the method, a Gamma-Gamma distribution parameter estimation process is simplified greatly through Mellin transform, and a uniform parameter estimation method is built specific to the problems of low fractional moment (FMOM) parameter estimation accuracy and even error estimation in the prior art. Then, further improvements are made, and a concept of double logarithmic cumulant expectation is put forward, so that a finial parameter estimation expression is deduced. Compared with an FMOM, the Gamma-Gamma distribution parameter estimation method has the advantages that relatively high computing efficiency is ensured, and the parameter estimation accuracy is increased.

The remote sensing image coarse positioning method based on Fourier-Mellin transformation comprises: (1) extracting short curve c and reference curve C for pre-processing before matching; (2) converting c into binary image named a; (3) with given step and rule, selecting all candidate curve ci and corresponding binary image bi from C; (4) matching bi with a in turn to calculate conversion parameters (alphai, si, x0i, and y0i); (5) taking inverse transformation to c by (alphai, si, x0i, and y0i) to obtain c'i, and calculating the Hi as Hausdorff distance between c'i and ci; (6) taking c* that meets Hmin=min{Hi} as the matching segment of c in C.

The invention discloses a Fourier-Mellin transform-based image geometric matching method, so as to solve problems of poor matching precision and long time due to a large rotation angle between a to-be-matched image and a reference image. The matching process comprises steps: the Fourier-Mellin transform is used for solving the rotation angle between the to-be-matched image and the reference image; the rotation angle is corrected to obtain an initial matching image; feature points of the above two image salient maps are extracted; the feature points are associated; an affine transform model is solved; and the model is used for transforming the initial matching image, a bilinear interpolation method is used for carrying out interpolation on the transformed image, and a final matching image is obtained. The condition of a small rotation angle difference between the to-be-matched image and the reference image can be effectively handled, the condition of a large rotation angle between the two images can also be processed, and the matching time is far less than that in an SIFT algorithm. The method of the invention is high in precision and matching efficiency, and can be used for processing image matching in the case of a large rotation angle difference between the images.

The invention discloses an image matching method based on multiscale Fourier-Mellin transform. The method comprises the following steps of acquiring a reference image I1 and an image to be matched I2 respectively, determining a Gaussian kernel at a pixel position (x, y) in the I2, and calculating a scale space image pixel value L(x, y, t) at the pixel position (x, y) in the I2; further acquiring a scale space image L(t) of the image to be matched I2, and carrying out bilinear interpolation on the L(t) respectively so as to acquire simulation imaging S(t) after the bilinear interpolation; setting a blank image whose size is the same with the size of the L(t), embedding a gray value of each pixel position in the S(t) into the blank image respectively so as to acquire a multiscale image, and calculating a Fourier-Mellin invariant image of the I1 and a Fourier-Mellin invariant image of the multiscale image S'(t) respectively; and calculating a phase correlation matching image between the scale space images L(t) of the I1 and the I2, calculating a convergent-divergent scale estimation value of a pixel position in the I1 and a rotation angle estimation value of the pixel position in the I1 respectively and then acquiring an optimal matching image of the I2.

The invention discloses a score Mellin transform-based optical image encryption method. The method comprises the following steps of: realizing score Mellin transform on an image according to an annular range in an optical mode; receiving a score Mellin transform result by using a charge coupled device (CCD) detector; taking a transform order of the score Mellin transform as a key; and taking a center coordinate of coordinate transform and the inner radius and the outer radius of an annular domain in which the transform is implemented as secondary keys. The method has high attack resistance and high safety.

The invention relates to the field of image matching technology, in particular to an image matching method and system. The image matching method comprises the steps that a, a template image and a to-be-matched image are read, and sparse Fourier transformation is performed on the template image and the to-be-matched image; b, an amplitude spectrum of the template image and an amplitude spectrum ofthe to-be-matched image after sparse Fourier transformation are calculated respectively; c, log-polar transformation is performed on the template image and the to-be-matched image according to the amplitude spectra; and d, phase correlation is performed on the log-polar transformation result of the template image and the to-be-matched image, a rotational angle, a scaling coefficient and a translation quantity are obtained, and matching is performed on the template image and the to-be-matched image according to the obtained rotational angle, scaling coefficient and translation quantity. Throughthe image matching method and system, a high-pass filtering step in traditional Fourier-Mellin transformation is omitted, an error brought by high-pass filtering in a traditional image matching algorithm is avoided, phase correlation done in a later period is more precise, and algorithm intelligence is obviously improved.

The invention discloses a fault traveling wave detection method based on VMD-S transform. The method comprises the steps of: S1: obtaining a fault traveling wave, and intercepting a traveling wave waveform of a certain time window, performing Karen Bell phase-mode transform, and extracting a line mode alpha component of the traveling wave; S2: performing variational modal decomposition (VMD) on the line mode alpha component to obtain K intrinsic mode functions (IMF), and then extracting the traveling wave waveform of an IMF1 component; S3: performing S transform on the IMF1 component to obtain a time-frequency distribution matrix of the waveform; S4: taking a module value of a matrix element to obtain an S mode matrix; and S5: extracting the first instantaneous frequency in the S mode transformation matrix obtained in the step S4, and calibrating a first amplitude sudden change point of the waveform to serve as an initial traveling wave front. The fault traveling wave detection methoddisclosed by the invention has the advantages of simple principle, high anti-interference ability, accurate calibration of the traveling wave head, and the like.

The invention relates to a method for encoding, decoding and measuring an encoding close-range photogrammetry marker. The method comprises the following steps: 1) encoding: 11) setting an encoding rate, 12) generating an encoding dictionary and 13) generating an encoding marker; 2) decoding: 21) preprocessing encoding images, 22) extracting a central point and 23) converting and decoding a one-dimensional barcode; and 3) measuring: after the ending of decoding, adopting an improved Fourier-Mellin transform algorithm for resolving the rotation scale relation and translation amount of locating marker points of encoding elements on two images. Compared with the prior art, the method provided by the invention has the advantages of fast decoding, controllable capacity, locating accuracy, capability of promoting rotating measurement accuracy, and the like.

An image matching system, method, and program for matching two images even if there is a difference of parallel movement, rotation angle, and scale between two images. A Fourier-Mellin transform is a Fourier transform and a log-polar coordinate transform to a registered image and a matching image. A phase only correlation unit and scalar information-rotation information generation unit generate correction information including scalar information and rotation information based on correlation strength of the phase information based on the transform result, a correction unit corrects the matching image RIM in accordance with the correction information, a parallel movement unit performs phase only correlation based on the image resulting from the correction processing and the registered image, a correlation value generation unit generates a correlation value based on the correlation image data, and a judgment unit performs judgment concerning the matching based on the correlation value.

The present invention provides a method and system for determining near-duplicate images. The method and system includes performing a Fourier-Mellin transform on each of a plurality of images. For each image of the plurality of images, the method and system includes generating a signature based on the Fourier-Mellin transform. The method and system includes comparing the signature of at least one of the images to at least one of the signatures of the other plurality of images and determining any near duplicate images based on the comparing of the signatures.

The invention provides a multi-image encryption method based on fractional mellin transform. The method includes the steps of compressing multiple images according to a frequency spectrum cutting and splicing method, then, dividing an image obtained through compression into a plurality of sub-images according to an annular domain, conducting fractional mellin transform on all the sub-images, encrypting a fractional mellin transform outcome into a cryptograph image through magnitude and phase encoding, taking transform degrees of fractional mellin transform as secret keys, and taking a center coordinate for coordinate transform, an inner radius of the annular domain for implementation of transform and an outer radius of the annular domain for implementation of transform as secondary secret keys. According to the multi-image encryption method, the multiple images are compressed through the real-value characteristics of fractional cosine transform, and therefore overweight transmission loads are avoided, and the requirement for instantaneity of image data transmission is met. Fractional mellin transform is then conducted on the image obtained through compression, the fractional order characteristic, the nonlinear characteristic and the like of the fractional cosine transform are fully utilized, and the multi-image encryption method has quite high robustness in terms of clipping prevention and noise prevention. The nonlinear operation is introduced into an encryption system while multi-image encryption transmission can be ensured, and safety of the encryption system is enhanced.

Disclosed is an image encryption method based on two-dimensional compressed sensing and non-linear fraction Mellin transform. Part of Hadamand matrix built through a chaotic system serves as a measurement matrix, projection measurement is carried out on an original image subjected to two-dimensional discrete cosine transform in two orthogonal directions to obtain a measurement value, compression encryption of the image is realized, and the defect of a too large data amount after fraction Mellin transform is carried out on the image is overcome; non-linear fraction Mellin transform is carried out on the measurement value according to an annular range to obtain a ciphertext, and the problem that a linear compressed sensing encryption system is low in safety is resolved. The initial value of the chaotic system and the transform order of fraction Mellin transform serve as the main secret key, and the central coordinates of coordinate transform and inner radius and outer radius of the transform implementation area serve as a secondary secret key. The two-dimensional compressed sensing technology and non-linear fraction Mellin transform are effectively combined, the consumption amount of the secret keys is small, the secret key space is large, and high safety is realized.

The invention relates to a narrow baseline parallax reconstruction method based on multi-scale superpixels and phase correlation. The method comprises the following steps: step 1, ; obtaining a globalsimilarity transformation model between the two input images by using a Fourier-Meilin transformation method; step 2, carrying out pixel polar error estimation on the global similarity transformationmodel through a multi-scale superpixel method to obtain a pixel polar error estimation result; step 3, for the pixel polar error estimation result, performing sub-pixel precision acquisition operation by adopting a sub-pixel phase correlation method based on singular value decomposition and random sampling consistency, and further obtaining a final narrow baseline parallax result; and step 4, performing three-dimensional information reconstruction on the image according to a final narrow baseline parallax result. Compared with the prior art, the method has the advantages of high precision, good robustness, low texture region matching efficiency and the like.

The invention is applicable to the technical field of transformer, and provides a transformer partial discharge fault diagnosis method and an online monitoring device, wherein a nonlinear mapping compression feature quantity method and a Merlin transform algorithm are adopted to extract a feature quantity from a partial discharge signal, and an eigenvector representing the characteristics of the partial discharge signal is formed, and mapped on a feature plane. The extracted feature quantity is clustered by unsupervised fuzzy c-means clustering algorithm to form a plurality of samples, and a phase and a waveform characteristic of each sample after clustering are analyzed, and relevant features are extracted. A sample fingerprint database is established by using a mathematical morphology particle analysis method to compare each sample, and a neural network algorithm is used to identify a partial discharge signal type pattern. According to the transformer partial discharge fault diagnosis method and the online monitoring device, the type of fault of partial discharge of the transformer is determined.

The invention discloses a target identification method and system based on a radar one-dimensional distance image. The echo data of a training sample pair is subjected to Fourier Mein transform processing, and then the echo data is input into a one-dimensional convolutional neural network model for training. The Fourier Meilin transformation has time shift invariance and scale invariance, and thescale invariance can enable the distance image to be insensitive to changes of azimuth angles and attitude angles in a Meilin transformation domain; Time shift invariance enables Fourier melin transformation to adapt to time translation and scale change of signals; Therefore, a stable feature vector can be obtained after the one-dimensional distance image is subjected to Fourier Mellin transformation, the attitude sensitivity, the translation sensitivity and the amplitude sensitivity of the distance image are reduced, the robustness of the distance image is enhanced, and the accuracy of targetidentification is improved.

The invention relates to a multi-image hiding method based on fractional mellin transform. According to the method, compression of multiple to-be-hidden images is carried out by utilizing a frequency spectrum cutting and splicing method, carrier images and compression images are respectively divided into multiple ring zone sub images, fractional mellin transform in different orders on each image is carried out, fusion of all results in a random combination mode is carried out by utilizing a primary Bezier curve, and fractional mellin transform is carried out by utilizing the same orders of carrier image transform to acquire fusion images containing to-be-hidden image information. The orders of fractional mellin transform and the fusion coefficients are taken as secret keys, and a center coordinate of coordinate transform and an inner radius and an outer radius for carrying out ring zone transform are taken as secondary secret keys. The method utilizes real property of discrete cosine transform to realize compression of the multiple images, increases hidden information amount in a frequency domain processing method, utilizes fractional mellin transform for encryption and improves anti-attacking capability and safety of systems.

The invention discloses a digital watermarking method based on digital holography and Fourier-Mellin transform. The method comprises the following steps that: carrying out conjugate symmetry extension and DFT (Discrete Fourier Transform) to obtain a two-dimensional discrete actual value function F (Mu, v), and mapping the actual value function to 0-255 to generate a digital watermarking hologram; in the intermediate frequency region of carrier image I DFT, embedding the rectangular block of the digital watermarking hologram, and embedding the same rectangular block into the centrosymmetric corresponding block of a magnitude spectrum; and carrying out inverse discrete Fourier transform on the new magnitude spectrum and an original phase spectrum to obtain a watermarking-contained image IW. An embedding method is simplified, embedding quality is improved, and the method achieves good balance between robustness and invisibility.

The invention provides an image registration method and system for train bottom fault detection of a motor train unit. The method comprises the following steps: acquiring an original image of the train bottom of the motor train unit; preprocessing the original image to obtain a first image; and performing image registration on the first image and the corresponding template image through Fourier-Mellin transform to obtain translation, rotation and scaling parameters, and processing the first image according to the obtained translation, rotation and scaling parameters to obtain a second image for fault detection. According to the method, all information in an image frequency domain is utilized instead of rigid transformation methods based on feature constraint points, such as affine transformation, so that high-precision registration of the train bottom image of the motor train unit is realized; the accurate position of the template image relative to the to-be-detected image is judged based on an image registration algorithm of Fourier-Mellin, and relative translation, rotation and scaling parameters are obtained.

The invention belongs to the field of signal processing, and provides a broadband nonlinear frequency modulation signal time difference / scale difference estimation method based on Fourier-Mellin transform for solving the time difference / scale difference estimation problem in a double-satellite passive positioning system. The core idea is to perform cascade estimation on time difference / scale difference by using Fourier transform time translation invariance and Mellin transform scale invariance. A result shows that compared with a traditional time difference / scale difference estimation method based on a broadband cross-ambiguity function, the method provided by the invention is lower in calculation complexity; and compared with a time difference / scale difference estimation method based on fractional Fourier transform, the method provided by the invention can be applied to linear frequency modulation (LFM),Quadratic Frequency Modulation (QFM), exponential frequency modulation (EFM), hyperbolic frequency modulation (HFM) and Gaussian noise pulse signals.

The invention discloses a cargo positioning method based on Fourier-Mellin transformation. The cargo positioning method comprises the following steps: step 1, establishing an absolute coordinate system, setting a calibration point, and identifying the position information of the calibration point positioning shelf; step 2, installing a camera at a dark place; step 3, establishing an image featurelibrary, extracting a shot specific frame image of the cargo as a feature image of a corresponding cargo position in real time, obtaining an affine transformation matrix through position information of the cargo shelf, and performing affine transformation to obtain a new feature image which is recorded as a cargo feature image library; step 4, obtaining a cargo image shot at the position of the current object taking structure, and obtaining a cargo image based on Fourier-Fourier transform; carrying out image registration of Meilin transformation, and calculating the pose relation between the goods and the goods taking mechanism; step 5, updating the feature graph library, and skipping to the step 4 to carry out subsequent image registration work. According to the method, high-precision positioning of various goods shelves and goods is achieved, calibrated image information is updated in time, the application range is wide, and operation is convenient.

The invention relates to a Gamma-Gamma distribution parameter estimation method based on a double logarithmic cumulant expectation, and belongs to the technical field of atmospheric turbulence channel parameter estimation of free space optical communication systems. According to the method, a Gamma-Gamma distribution parameter estimation process is simplified greatly through Mellin transform, and a uniform parameter estimation method is built specific to the problems of low fractional moment (FMOM) parameter estimation accuracy and even error estimation in the prior art. Then, further improvements are made, and a concept of double logarithmic cumulant expectation is put forward, so that a finial parameter estimation expression is deduced. Compared with an FMOM, the Gamma-Gamma distribution parameter estimation method has the advantages that relatively high computing efficiency is ensured, and the parameter estimation accuracy is increased.

The invention provides a multi-image encryption method based on fractional mellin transform. The method includes the steps of compressing multiple images according to a frequency spectrum cutting and splicing method, then, dividing an image obtained through compression into a plurality of sub-images according to an annular domain, conducting fractional mellin transform on all the sub-images, encrypting a fractional mellin transform outcome into a cryptograph image through magnitude and phase encoding, taking transform degrees of fractional mellin transform as secret keys, and taking a center coordinate for coordinate transform, an inner radius of the annular domain for implementation of transform and an outer radius of the annular domain for implementation of transform as secondary secret keys. According to the multi-image encryption method, the multiple images are compressed through the real-value characteristics of fractional cosine transform, and therefore overweight transmission loads are avoided, and the requirement for instantaneity of image data transmission is met. Fractional mellin transform is then conducted on the image obtained through compression, the fractional order characteristic, the nonlinear characteristic and the like of the fractional cosine transform are fully utilized, and the multi-image encryption method has quite high robustness in terms of clipping prevention and noise prevention. The nonlinear operation is introduced into an encryption system while multi-image encryption transmission can be ensured, and safety of the encryption system is enhanced.

A liquid foreign matter visual detection method comprises the following steps of performing image registration based on Fourier-Mellin transform and an unsupervised deformation registration model; preprocessing the registered image; performing image sequence frame difference on the preprocessed image; post-processing the image after the sequence frame difference; obtaining an adaptive deep learning model; and extracting system flaws by a generalized multi-image matting algorithm. The invention further provides a liquid foreign matter visual detection system, and the precision of image registration is improved. A tiny foreign matter segmentation result is clearer, and background noise is further filtered; the detection precision is ensured, the production benefit is increased, and the production loss is reduced.

The invention discloses a geometric fine correction method suitable for a high-resolution remote sensing image, relates to the field of geometric fine correction of remote sensing images, and solves the problem that in the prior art, an image matching method is easy to have wrong matching, so that geometric fine correction fails and the like. An image retrieval method in the field of computer vision is applied to geometric fine correction of remote sensing images, redundant calculation in the homonymy point matching process is reduced, and the running speed is remarkably increased; fourier-Mellin transform blocks are used for matching homonymy points of the images, parameters and thresholds suitable for high-resolution remote sensing images are provided through a large number of experiments, and a more stable and accurate result is obtained under the condition that ground feature change, radiation and color difference exist between the to-be-processed image and a reference base map. And full-automatic geometric fine correction is carried out by taking the data as a control point. The method provided by the invention is remarkably improved in the aspects of speed, precision and stability.

The invention provides a target tracking method based on Fourier-Merlin transform, and belongs to the field of computer vision and digital image processing. The problems of rotation and scale change of a tracked target in a target tracking technology are solved; the method comprises the following specific steps: reading a to-be-tracked video frame sequence, selecting a to-be-tracked target in a first frame mark, performing feature extraction on a search region, converting the search detection region from a Cartesian coordinate system to a logarithmic polar coordinate system, performing rotation and scale transformation on the target under the Cartesian coordinate system, and further converting the target into translation transformation under the logarithmic polar coordinate system; employing fourier-Merlin transform to obtain a tracking target rotation angle and a scale transform size; and carrying out interpolation updating on the tracked target template. According to the method, highrobustness and long-time real-time tracking can be realized, and when the target rotates or the scale changes, the target can be tracked in time.