30 results about "Log Gabor filter" patented technology

Disclosed is a high-speed train driver alertness detecting method based on face image and eye movement analysis. The method includes steps of acquiring images of the head of a driver by a primary camera and a secondary camera simultaneously, extracting characteristics of the head of the driver step by step by the aid of wavelet multi-scale, recognizing the head of the driver via a neural network training method, and detecting the face of the driver via an AdaBoost arithmetic; respectively starting a Harr characteristic and a two-dimensional orthogonal Log-Gabor filtering phase characteristic to detect human eyes under different illuminations; building an adaptive non-linear level-log model of the driver; realizing eye movement tracking analysis for the driver via a strong tracking particle filter arithmetic; and finally realizing weight sum for an eye movement characteristic fatigue factor value and six face images including PERCLOS, pupilla, blinking, nodding, yawning and face inclination so as to obtain an alertness value of the driver. The calculated alertness value is high in accuracy and robustness.

The invention discloses a fingerprint identification method. A modified fingerprint partitioning algorithm is provided in order to overcome the defects of a variance method used for carrying out fingerprint partitioning, an appropriate Log Gabor filter is built to achieve the reinforcement of a fingerprint image, and finally a node relationship representation method under a curvilinear coordinate system is provided. The modified fingerprint partitioning algorithm is provided in order to overcome the defects of the variance method used for carrying out fingerprint partitioning, and the fingerprint partitioning algorithm makes full use of the advantages of a simple algorithm and high calculation speed of the variance method, overcomes the defect that the variance method is easily affected by noise, and obtains a more accurate fingerprint portioning image. The appropriate Log Gabor filter is built to achieve the reinforcement of the fingerprint image, and the node relationship representation method under the curvilinear coordinate system can be effectively matched with deformed fingerprints. The system distinguishes the identification of each person by utilizing the characteristics such as uniqueness and forgery difficulty of the fingerprints, and has the advantages that allograph is avoided and paper is not used.

The invention discloses a high-precision palmprint recognition method based on two-dimensional orthogonal Log-Gabor filtration, which uses two two-dimensional Log-Gabor filters which are mutually orthogonal on directions to constitute the two-dimensional orthogonal Log-Gabor filter OLG (f, Theta, Theta 0, Sigma Theta, f0, Sigma f), thus carrying out the filtration of a palmprint image after the pre-treatment; then the palmprint image OF after the filtration adopts the phase coding mode to extract the orthogonal phase characteristic POPF which is equal to (FR, FI) of the palmprint image; then the Hamming distance matching algorithm is adopted to carry out the matching of orthogonal phase characteristics POPF1 and POPF2 of two palmprint images, thus obtaining the matching result. The algorithm is less affected by the conditions of the collected light and has low computational complexity and high recognition precision.

The invention discloses a 3D image quality evaluation method based on visual saliency and depth map. The method includes: obtaining disparity map by stereo matching algorithm, Using spectral residualalgorithm to obtain the saliency map of image, using structural similarity algorithm to obtain the similarity of brightness, contrast and contrast, and using Gaussian color model to obtain chromaticity similarity characteristics and other metrics, finally, using Log Gabor filter to extract multi-scale features from reference image and distorted image, obtaining edge texture features of left and right images in multi-scales and multi-directions, carrying out the similarity calculation, obtaining the index features of image quality evaluation, obtaining objective quality fractions by regressionprediction with support vector machine, completing stereo image quality mapping, and obtaining final stereo image quality evaluation. The objective evaluation and the subjective evaluation of the full-reference image quality evaluation method provided by the invention have good consistency, and are superior to the traditional stereoscopic image quality evaluation method.

The invention relates to an image retrieval and classification method through a Log-Gabor filter. The method mainly comprises the following steps of 1) utilizing a copula model to capture correlation of Log-Gabor wavelet sub-bands and correlation of a color channel to improve retrieval and classification accuracy, and 2) selecting the Log-Gabor wavelet sub-bands (eliminating redundant sub-bands to improve algorithm precision and reduce calculation time) on the color channel, building one copula model for each color picture to improve calculation efficiency. Compared with the prior art, the color picture retrieval and classification method achieves strong noise interference resistance, great retrieval and classification performance and low calculation complexity.

The present invention provides an image quality evaluation method and system, and belongs to the technical field of image processing. The method comprises: performing 5-dimensional decomposition on a reference image and a distorted image separately by using a Log-Gabor filter, and then performing Riesz transform on a decomposed image based on a theory of single-speech signal, to obtain 5 pairs of single-speech signals; calculating a similarity degree of local features of the 5 pairs of single-speech signals to obtain 5 similarity maps, and performing weighted summation by using a characteristic of a human visual system to obtain 1 single-speech similarity map; In addition, performing calculation of single-speech phase consistency on a Riesz transform feature graph obtained by using the reference image, to obtain a pooling function; and calculating a gradient magnitude similarity degree for the reference image and the distorted image, and finally performing convolution on a product of the single-speech similarity degree and the gradient magnitude similarity degree by using the pooling function, to obtain a final image quality evaluation index. The algorithm is tested on multiple databases, and as proven from results, the RGSIM algorithm has excellent overall performance.

The invention discloses a method for realizing image registration of a synthetic aperture radar (SAR) by using three components of monogenic signals, belonging to the technical field of image processing and registration of SARs. The conventional gradient information-based detection method and cross correlation matching method have the defects of excessive detected characteristic points, overlong detection time, low registration accuracies and the like when applied to image registration of an SAR. A method for matching a detection algorithm of monogenic signal phase congruency with monogenic signal phase correlation is given, and three orthogonal monogenic signal component signals are generated by designing a frequency domain Log-Gabor filter. One path of the monogenic signal component signals is transmitted to a characteristic detector, i.e., a local amplitude and a local phase are resolved by using three components to construct a monogenic signal phase congruency function for detecting the phase congruency characteristic. Another path of the monogenic signal component signals is transmitted to a matcher, and a characteristic description vector is constructed by using three components of the monogenic signals; a characteristic vector correlation matrix is obtained by calculating the characteristic vector correlation of a reference image and characteristic points in an image to be registered; and largest line elements and column elements in the characteristic vector correlation matrix are searched and are indexed to a coarsely-matched characteristic point pair, so that coarse characteristic matching is realized. An affine basic matrix is fitted by using an RANSAC (Random Sample Consensus Algorithm), so that accurate matching of characteristic points is completed. An affine conversion model is used for realizing image registration of the SAR. The algorithm disclosed by the invention has the advantages of realization of automatic registration of SAR images, high registration speed, small influence by speckle noise, high registration accuracy and popularization and application values.

The invention provides an improved iris recognition method and aims to improve the accuracy of iris recognition and shorten the time of iris recognition. The method includes decomposing an iris image through wavelet transform to obtain low frequency sub-band information of the iris image, extracting texture features of the iris by utilizing two-dimensional log-Gabor filter banks in different directions and dimensions, and matching generated iris feature codes by utilizing a support vector machine. The iris recognition method is more suitable for real-time iris recognition.

The invention provides a texture image segmentation method based on a level set model. The texture image segmentation method comprises following steps of: 1, carrying out Log-Gabor filtering on an input image and outputting the filtered image; 2, extracting an LSS descriptor from the filtered image so as to obtain an image formed by a descriptor component; 3, constructing a texture energy item on the image formed by the descriptor component; 4, constructing an LGDF energy item on the input image; 5, integrating the two kinds of energy items in steps 3 and 4 so as to form an energy function of the method; and 6, by maximizing the energy function in the step 5, solving a segmentation curve of the energy function and using a Lattice Boltzmann method to accelerating the solving process. According to the invention, based on the level set model, a new texture energy item is designed by use of the LSS descriptor and then is combined with a traditional LGDF energy item, a good segmentation effect is achieved for each kind of texture image, so the method has robustness for noise and uneven gray levels, and through optimization of the Lattice Boltzmann method, the image segmentation efficiency is improved.

The invention discloses a no-reference stereo image quality evaluation method based on dictionary learning and machine learning. The method comprises the following steps of: firstly, performing log-Gabor filtering for left and right viewpoint images, obtaining respective amplitude and phase information, then, performing local binarization operation for the amplitude and the phase information, and obtaining a local binarization mode feature image of the left and right viewpoint images; secondly, using a binocular energy model to fuse the amplitude and the phase information of the left and right viewpoint images, obtaining binocular energy information, and acquiring a local binarization mode feature image of the binocular energy information; then, using a coordination representation algorithm to perform dictionary learning for the local binarization mode feature images of the left and right viewpoint images and the binocular energy information, obtaining binocular visual perception sparse feature information, and finally, obtaining an objective quality evaluation predicted value of a to-be-evaluated distorted stereo image. The method has the advantages of being capable of fully considering stereo visual perception characteristics, and being capable of effectively improving correlation between objective evaluation result and subjective perception.

The invention relates to a local feature description method, system and device for rotation invariant multi-source images. According to the method, a descriptor for local feature matching of the rotation invariant multi-source images is constructed based on a multi-scale multi-direction log Gabor filter; and the descriptor not only can adapt to difference between the multi-source images caused bynonlinear brightness change but also has a certain suppression effect on noises, and can adapt to geometric changes between the images, such as translation, rotation and the like, so that the descriptor has stronger robustness, and the descriptor can be more accurate and efficient for the local feature description method supporting rotation invariance.

A method and apparatus of identification by extracting and matching biometric detail from a subcutaneous vein infrared image. The image's Region of Interest is identified and artifacts are removed. A bank of filters, such as Symmetric Gabor Filters, Complex Gabor Filters, Log Gabor Filters, Oriented Gaussian Functions, or Wavelets, filters the image into a set of key value images that are subdivided into regions. An enrollment key, defined by ordered statistical measures of pixel intensities within the regions, is compared using a distance metric to a stored verification key. Various statistical measures may be used, such as variance, standard deviation, mean, absolute average deviation, max value, min value, max absolute value, median value, or a combination of these statistical measures. Various distance metrics may be used, such as Euclidean, Hamming, Euclidean Squared, Manhattan, Pearson Correlation, Pearson Squared Correlation, Chebychev, or Spearman Rank Correlation.

The invention provides an image quality evaluation method based on log Gabor filter phase similarity. According to the method, first, 18 pairs of log Gabor filters are adopted to make a similarity comparison between an original image and a to-be-detected image, then, phase information is extracted; and finally, the quality of the to-be-detected image is obtained by comparing the average cross correlation of phase diagrams of the two images. The method of the invention has the advantages of high accuracy, low complexity, and high processing speed.

The invention discloses a high-resolution remote sensing image segmentation method combining a phase and a spectrum. For the limitation existing in the aspects of weak edge detection and the like dueto the fact that gradient information is excessively depended during segmentation of a traditional method, advantage complementation is conducted by introducing phase information, and firstly, phase consistency information is extracted through Log Gabor filtering; on the basis, a local homogeneity index J-value in a traditional JSEG algorithm is adopted to optimize an edge detection result, and anobjective function optimization strategy based on inter-scale mutual information minimization is provided, so that a parameter adaptive optimal parameter combination and response model is constructed, and finally, the multi-scale region segmentation and region merging is carried out based on the model to obtain a final segmentation result.

The invention relates to a method for recognizing iris on the basis of odd-symmetry 2D (two-dimensional) Log-Gabor filter and Adaboost combinations. The method has the advantages that the problem of partial non-effective feature points when iris features are extracted by 2D Log-Gabor filters can be solved by odd-symmetry 2D Log-Gabor filters, phase and amplitude signals of iris textures are analyzed at different frequencies in different directions, redundant information is eliminated by the aid of feature fusion technologies, accordingly, feature codes can be formed and can be trained by the aid of Adaboost algorithms to form the optimal classifiers, and features can be recognized.

The invention relates to an image strong and weak edge detection method based on spatio-temporal information responded by dot matrix nerve cells. Firstly, a multi-direction Log-Gabor filtering result of an image in a view is utilized and edge information of the image is reconstructed; a reconstruction result serves as input of the dot matrix nerve cells; time in releasing action potentials of all the nerve cells is recorded to form a time matrix; a constructed reception field window slides on the time matrix, the improved variance is calculated according to a time order of all time elements and a center point of the window undergoes assignment and thus, a variance matrix containing nerve cell responding time and the spatio-temporal information is obtained; afterwards, the reception field window continues sliding on the variance matrix, the side direction rejection characteristic of the nerve cells in space is achieved and an edge matrix is obtained; finally, the edge matrix is mapped into a result image inversely. According to the image strong and weak edge detection method based on the spatio-temporal information responded by the dot matrix nerve cells, the spatio-temporal information responded by the dot matrix nerve cells is taken into consideration, the edges of the image can be detected and additionally, a strong and week relationship of the edges can be effectively reflected.

The invention discloses a universal no-reference image quality evaluation method based on a phase selection mechanism. The method comprises the steps of firstly implementing Log-Gabor filtering on a distorted image to be evaluated, thus obtaining a multi-scale and multi-directional phase image; comparing the pixel value of each pixel point in the phase image with the pixel values of surrounding pixel points to obtain a local feature map; then obtaining the local feature mode chart of the local feature map by a rotational invariance method, and performing statistics on the local feature mode chart by a histogram statistic method, thus obtaining the histogram statistic feature vector of the distorted image to be evaluated; and at last obtaining the objective quality evaluation predicted value of the distorted image to be evaluated according to the distance between the histogram statistic feature vector of the distorted image to be evaluated and the histogram statistic feature vector of each distorted image in a training set. The method has the advantages that the influence of the change of the phase information on the visual quality can be fully considered, and the correlation between the objective evaluation result and subjective perception can be effectively improved.

The invention relates to a 2D Log Gabor and composite coding method-based infrared iris authentication system and a method. The authentication system comprises an infrared iris image acquisition module used for acquiring a to-be-recognized iris image; an iris feature extraction module used for extracting iris two-dimensional texture features through a 2D Log Gabor filter; an iris feature vector encoding module used for encoding extracted feature vectors through a composite coding method; and an iris feature code matching module used for carrying out feature classification and recognition by using a projection extreme learning machine as a classifier. The authentication method comprises the following steps of acquiring a to-be-recognized iris image and preprocessing the image; extracting iris texture features; encoding the feature vectors of the extracted iris texture features; classifying the features by adopting a projection extreme learning machine as a classifier; calculating the hamming distance according to the classification result; and judging whether irises are from the same person or not according to the hamming distance. The system and the method are high in authentication precision.

The invention provides a filter bank-based gait phase recognition method for children suffering from cerebral palsy. The method comprises the stages of original signal obtaining, wherein original signals of surface electromyogram of the children suffering from the cerebral palsy are obtained; preprocessing, wherein the received original signals are preprocessed; feature library establishing, wherein on the basis of a log-gabor filter bank, surface electromyogram signals obtained after preprocessing are subjected to multi-band filtering, and a feature library is established by conducting feature extraction on the basis of linear discriminant analysis; classification, wherein an extreme learning machine is adopted as a classifier, feature subsets picked from the feature library and corresponding label information serve as training samples, and ELM modeling is conducted, so that gait phase classification in the normal walking process of the children suffering from the cerebral palsy is evaluated.

The invention discloses an underwater bubble image feature segmentation and extraction method, and the method comprises the steps: firstly obtaining the central frequency of an image, and guaranteeingthat a Log Gabor filter bank covers the main frequency components of the image, and can extract the edge information of an input image more effectively; then, performing edge detection on the image,simplifying the edge detection of the image into local energy detection of the image, obtaining extreme values globally through normalization processing of energy of a local area of the image, and preventing respective solving of the local extreme values so that the influence of illumination intensity on the edge detection process of the image is small; removing short and small boundaries, dividing the image into simple geometric structures, and performing region growth by taking pixels around the divided boundaries as seed points; and finally, filling small holes of the image, and completingfeature segmentation and extraction of bubbles in the underwater image. By adopting the method provided by the invention, the bubble contour features in the underwater image can be accurately extracted, so that the number of bubbles in water is counted and calculated, and the real-time monitoring of seawater bubbles is supported.

The invention discloses a no-reference stereo image quality evaluation method based on dictionary learning and machine learning. The method comprises the following steps of: firstly, performing log-Gabor filtering for left and right viewpoint images, obtaining respective amplitude and phase information, then, performing local binarization operation for the amplitude and the phase information, and obtaining a local binarization mode feature image of the left and right viewpoint images; secondly, using a binocular energy model to fuse the amplitude and the phase information of the left and right viewpoint images, obtaining binocular energy information, and acquiring a local binarization mode feature image of the binocular energy information; then, using a coordination representation algorithm to perform dictionary learning for the local binarization mode feature images of the left and right viewpoint images and the binocular energy information, obtaining binocular visual perception sparse feature information, and finally, obtaining an objective quality evaluation predicted value of a to-be-evaluated distorted stereo image. The method has the advantages of being capable of fully considering stereo visual perception characteristics, and being capable of effectively improving correlation between objective evaluation result and subjective perception.

The invention discloses a high-resolution remote sensing image segmentation method combining a phase and a spectrum. For the limitation existing in the aspects of weak edge detection and the like dueto the fact that gradient information is excessively depended during segmentation of a traditional method, advantage complementation is conducted by introducing phase information, and firstly, phase consistency information is extracted through Log Gabor filtering; on the basis, a local homogeneity index J-value in a traditional JSEG algorithm is adopted to optimize an edge detection result, and anobjective function optimization strategy based on inter-scale mutual information minimization is provided, so that a parameter adaptive optimal parameter combination and response model is constructed, and finally, the multi-scale region segmentation and region merging is carried out based on the model to obtain a final segmentation result.

The invention discloses a 3D image quality evaluation method based on visual saliency and depth map. Use the stereo matching algorithm to obtain the disparity map; use the spectral residual algorithm to obtain the saliency map of the image, use the structural similarity algorithm to obtain the similarity of brightness, contrast, and contrast, and use the Gaussian color model to obtain metrics such as chroma similarity features; Finally, the Log Gabor filter is used to extract multi-scale features from the reference image and the distorted image, and the edge texture features of the left and right images of multi-scale and multi-directional are obtained, and the similarity calculation is performed to obtain the index features of image quality evaluation, and the support vector machine is used to carry out Regression prediction, get the objective quality score, complete the mapping to the stereoscopic image quality, and get the final stereoscopic image quality evaluation. The objective evaluation and subjective evaluation of the full-reference image quality evaluation method proposed by the present invention have good consistency, and are superior to the traditional three-dimensional image quality evaluation method.

The invention relates to a local feature description method, system and device of a rotation-invariant multi-source image. This method is based on multi-scale and multi-directional log Gabor filters to construct a descriptor for local feature matching of rotation-invariant multi-source images. This descriptor can not only adapt to the differences between multi-source images caused by nonlinear brightness changes, but also has certain noise At the same time, it can also adapt to geometric changes between images, such as translation and rotation, so as to ensure that the descriptor has stronger robustness, making the descriptor more accurate for local feature description methods that support rotation invariance and efficient.

The invention discloses a high-precision palmprint recognition method based on two-dimensional orthogonal Log-Gabor filtration, which uses two two-dimensional Log-Gabor filters which are mutually orthogonal on directions to constitute the two-dimensional orthogonal Log-Gabor filter OLG (f, Theta, Theta 0, Sigma Theta, f0, Sigma f), thus carrying out the filtration of a palmprint image after the pre-treatment; then the palmprint image OF after the filtration adopts the phase coding mode to extract the orthogonal phase characteristic POPF which is equal to (FR, FI) of the palmprint image; then the Hamming distance matching algorithm is adopted to carry out the matching of orthogonal phase characteristics POPF1 and POPF2 of two palmprint images, thus obtaining the matching result. The algorithm is less affected by the conditions of the collected light and has low computational complexity and high recognition precision.

The invention discloses a face recognition system based on a manifold learning subspace algorithm. The system comprises a data analysis and processing center, the data analysis and processing center comprises a data processing module, an image comparison module and a data receiving module, the data processing module comprises an irrelevant linear discriminant embedding module and an orthogonal unsupervised discriminant mapping module, and a server is arranged at the output end of the data analysis and processing center. According to the invention, the filter is arranged, and the filter is setto a Log-Gabor filter and can transfer a function as a high function on a logarithmic frequency scale. the frequency response of the original image can be reflected more truly; the filter can have nodirect component, the bandwidth is not limited, the filter with any bandwidth and zero DC component can be constructed, the natural image can be better represented, the filter works to expand the feature dimension of the video image, a data processing center can conveniently carry out dimension reduction calculation on the video image, and a more accurate result is obtained.

The invention discloses a method for realizing image registration of a synthetic aperture radar (SAR) by using three components of monogenic signals, belonging to the technical field of image processing and registration of SARs. The conventional gradient information-based detection method and cross correlation matching method have the defects of excessive detected characteristic points, overlong detection time, low registration accuracies and the like when applied to image registration of an SAR. A method for matching a detection algorithm of monogenic signal phase congruency with monogenic signal phase correlation is given, and three orthogonal monogenic signal component signals are generated by designing a frequency domain Log-Gabor filter. One path of the monogenic signal component signals is transmitted to a characteristic detector, i.e., a local amplitude and a local phase are resolved by using three components to construct a monogenic signal phase congruency function for detecting the phase congruency characteristic. Another path of the monogenic signal component signals is transmitted to a matcher, and a characteristic description vector is constructed by using three components of the monogenic signals; a characteristic vector correlation matrix is obtained by calculating the characteristic vector correlation of a reference image and characteristic points in an image to be registered; and largest line elements and column elements in the characteristic vector correlation matrix are searched and are indexed to a coarsely-matched characteristic point pair, so that coarse characteristic matching is realized. An affine basic matrix is fitted by using an RANSAC (Random Sample Consensus Algorithm), so that accurate matching of characteristic points is completed. An affine conversion model is used for realizing image registration of the SAR. The algorithm disclosed by the invention has the advantages of realization of automatic registration of SAR images, high registration speed, small influence by speckle noise, high registration accuracy and popularization and application values.

The invention discloses a universal no-reference image quality evaluation method based on a phase selection mechanism. The method comprises the steps of firstly implementing Log-Gabor filtering on a distorted image to be evaluated, thus obtaining a multi-scale and multi-directional phase image; comparing the pixel value of each pixel point in the phase image with the pixel values of surrounding pixel points to obtain a local feature map; then obtaining the local feature mode chart of the local feature map by a rotational invariance method, and performing statistics on the local feature mode chart by a histogram statistic method, thus obtaining the histogram statistic feature vector of the distorted image to be evaluated; and at last obtaining the objective quality evaluation predicted value of the distorted image to be evaluated according to the distance between the histogram statistic feature vector of the distorted image to be evaluated and the histogram statistic feature vector of each distorted image in a training set. The method has the advantages that the influence of the change of the phase information on the visual quality can be fully considered, and the correlation between the objective evaluation result and subjective perception can be effectively improved.