最新刊期
卷 21 ， 期 8 ， 2016
- 摘要：A single image super-resolution reconstruction method based on a variational calculus model for multi-direction stencils is proposed in this study to enhance the accuracy of image super-resolution reconstruction and avoid distortion on the edges and areas with a rich texture caused by few stencil directions in classical interpolation. First, multi-direction stencils are built to reflect 28 directions used to calculate the contour directions of an input image. Then, the TV model is used to estimate image contours and to determine the optimized direction of the edge contours. Finally, image super-resolution reconstruction is realized through image interpolation based on the aforementioned multi-direction stencils. Compared with those in the classical high-resolution test image reconstruction method of image interpolation with contour stencils, the average peak signal-to-noise ratio and the average structural similarity obtained using the method presented in the current work are increased by 1.578 dB and 0.030 02 dB, respectively. The method proposed in the current work can effectively overcome the negative effects of traditional interpolation methods, such as blurred and jagged edges of images. Furthermore, this method can avoid distortion on the edges and areas with a rich texture caused by a few stencil directions. The effectiveness of this method has been verified via a large number of simulations.关键词：multi-direction stencils;super resolution reconstruction of image;sub-pixel point;variational calculus model;interpolation3203|394|1更新时间：2024-05-07
- 摘要：The learning-based methods for single image super-resolution employ an instance training database to produce a high-resolution (HR) image from a single low-resolution (LR) input. In this study, we propose a new super-resolution framework without external training database. The proposed method is based on the self-similarity of images, which is a recurrence of image patches within an image or across image scales, and the support vector regression (SVR) model derives good fitting data via nonlinear mapping. First, the image pyramid of the input LR images is established, and the set of LR/HR image patch pairs is set. Second, we search similar image patches of input LR image patch in the set of LR/HR image patch pairs. Then, we use SVR to learn the map relationship between these similarity LR image patches and the pixel value of the center of the corresponding HR images. Finally, we can obtain the HR image patch through the aforementioned relationship and input LR image patch. We tested the proposed method on seven HR images with different textures and structures, which are downsampled by Gaussian blurring under a scalar factor of 2. The average PSNRs of our method are 2.37, 0.70, and 0.57 dB higher than the bicubic interpolation, the sparse representation-based super-resolution method, and the support vector regression-based super-resolution method, respectively. Experimental results show that the proposed method can effectively achieve image super-resolution reconstruction, particularly for the image with a highly similar texture.关键词：single-image;super-resolution;self-similarity;support vector regression;pyramid2708|309|4更新时间：2024-05-07
- 摘要：Due to the problem that image inpainting method based on the exemplar easily leads to sawtooth in the filled frontier and because the stage-wise inpainting strategy does not provide sufficient constraints on the priority of patch-filling, we propose an inpainting method based on arc promoting, exemplar searching and priority computation. Firstly, we carry out the mean filtering operation to calculate the isophotes more precisely. Second, we add the local feature information to further restrict the definition of priority and make the patch-filling order more reasonable. Third, we measure the candidate exemplar patches by adding gradient information, which can decrease the number of candidate patches. Fourth, we search the exemplar patches hierarchically to balance time efficiency and space efficiency. Finally, we fill the target region using an arc promoting method to keep the edge smooth. The inpainting result analysis of our method is contrasted with the analysis based on other methods. Subjectively, the method can maintain visual connectivity. Objectively, the value of the peak signal-to-noise ratio (PSNR) were higher than other inpainting methods. Not noly can the method be used to repair the images of natural and cultural relics, but also it has good application in the removal of the object. The repair effect is good, and the applicability is strong.关键词：arc promoting;exemplar searching;priority definition;image inpainting2857|346|8更新时间：2024-05-07
- 摘要：The artistic dot which has microscopic security features is designed based on threshold matrices screening method, then the hidden information is loaded into the plate incorporating microstructure dots using halftone information hiding technology, and research its potential application value. Halftone processing to get halftone images, the plate loaded hidden information is screened by a modified threshold matrix. Then, the information is embedded into the plate incorporating microstructure dots, when the revealing layer with matching parameters is superposed on it the hidden information can be realized, and the quality of information hiding and extraction is evaluated. Through the information hiding and extraction experiment, the subjective visual evaluation results show that the information hiding effect on microstructure dot achieved good visual performance both in single channel and multiple channels. The same conclusion was obtained by quantitative evaluation, which reached over 0.995 of SSIM. Furthermore, information extracting achieved a clear and complete effect. The security standard, efficiency, and added value of print and package products are improved with nearly no increase in cost. In particular, the proposed technique has extensive application prospects and value in the counterfeiting of security documents, such as banknotes, checks, certificates, and travel documents, as well as in valuable products, such as the packages of electronic image publishing products, medical drugs, and food products.关键词：halftone;threshold matrix;micro-structure dot;information hiding;anti-counterfeiting2709|233|1更新时间：2024-05-07
- 摘要：To solve the problems of low contrast and brightness as well as high noise level in low-illumination videos, a fast and effective low-illumination video enhancement algorithm is proposed by combining retinex theory with dark channel prior theory to improve contrast and reduce noise. Considering enhancing low illumination videos and amplifying noise simultaneously, removing noise before enhancing videos is beneficial to improving video enhancement effects. Therefore, this study combines the advantages of guided filtering and median filtering to propose an improved comprehensive denoising algorithm, which is applied to the YCbCr space. Then, the luminance transmission map is estimated by extracting luminance components. Furthermore, the atmospheric model is applied to recover the low-illumination video. Finally, scene detection, edge compensation, and inter-frame compensation are added to further improve the effectiveness and speed of the process. The proposed algorithm can effectively improve the brightness and contrast of low-illumination videos, reduce noise, strengthen the detailed information of videos, and diminish video scintillation, thereby improving the quality of videos. The proposed algorithm has a dominant advantage in processing speed, which is over 10 times faster than Dong's algorithm and the Retinex algorithm. Experimental results show that the proposed algorithm exhibit superior performance over other algorithms. First, the continuity of inter-frame motion can be guaranteed. Second, the enhancement effects and processing speed can be improved. Third, details and edging outlines are processed carefully, which results in unique effects that cannot be achieved by other algorithms. Therefore, the proposed algorithm can be applied in various areas, such as video surveillance, target tracking, and intelligent transportation systems, to achieve real-time video enhancement.关键词：low-illumination video enhancement;atmospheric physical model;luminance transmission map;noise reduction;inter-frame processing2938|351|3更新时间：2024-05-07
- 摘要：Image set matching has attracted increasing attention in the field of pattern recognition. For set-based image matching, the key issues can be categorized on the basis of the processes of representing the image set and measuring the similarity between two sets. Support vector domain description (SVDD) is a recently developed method based on support vector machine learning. SVDD is a boundary one-class learning method that maximizes the availability of samples that do not belong to the target class in refining its decision boundary, and can be used to describe a set of objects. Accordingly, each image set is described with a hypersphere, and the problem of image set matching is converted into the measure of the distance between two hyperspheres. Using support vector machine learning, each image set from the original input space is mapped into a high-dimensional feature space and modeled with support vector domain to handle the underlying non-linearity in the data space. In the feature space, a hypersphere encloses most of the mapped data. Thereafter, a novel metric is proposed based on domain–domain distance in a high-dimensional feature space; the distance between two image sets is then converted into the distance between pair-wise domains. However, the SVDD model has a disadvantageously simple form with only a single kernel information. Selecting the best kernel parameters is difficult and the constructed hypersphere is considerably sensitive to the trade-off parameter. Multiple kernel learning methods apply multiple kernels instead of merely one specific kernel function and its corresponding parameters. Recent developments in composition kernel learning for classification motivated us to apply a position-based weighting instead of the same global trade-off parameter to discriminate the importance of samples. Furthermore, considering the SVDD model's disadvantageously simple form with only one kernel and the difficulty of selecting the best kernel parameters, we propose a multi-kernel SVDD model, which can flexibly describe the data distribution boundary in the feature space after analyzing the space of multi-kernel mapping. This study utilizes the nearest neighbor classifier to obtain the class label. This study's experimental settings reach 100%, 98.72%, and 62.34% recognition rate in the public Honda/UCSD, CMU MoBo, and YouTube video database, respectively. Given that multi-kernel learning can improve the efficiency of kernel selection and automatically evaluate the relative importance of the candidate kernels, the multi-kernel SVDD model flexibly describes the data distribution boundary in the feature space and provides a considerably accurate data description for the multifaceted context of the multi-model data set. Experiments conducted on public data sets demonstrate that the multi-kernel SVDD improves prediction accuracy and assists in characterizing the properties of complex data.关键词：set matching;pattern recognition;face recognition;support vector domain description;distance measure;multiple kernel learning2436|206|0更新时间：2024-05-07
- 摘要：The edge of an image holds important visual information, which plays an important role in the subsequent image understanding and scene perception. Edge detection is used to extract image edge information and eliminate irrelevant information, which significantly reduce the amount of data for the subsequent analysis. Numerous scholars have proposed various edge operators for different requirements. However, the detection accuracy of traditional edge detectors is slightly low when extracting edges. Furthermore, the texture details in an image, which include complex texture with a fractal structure and weak edges, and noise immunity capability, are also weak. A fractional order derivative has the advantage of strengthening and extracting the textural features and weak edges of digital images. A person can choose a different order for the fractional calculus according to various images and interesting features to improve high-frequency signals, nonlinearly enhance intermediate-frequency signals, and retain low-frequency signals. To address the aforementioned problem, this study proposes an improved Canny edge detector based on a fractional order derivative. The method calculates image gradient using the classical Grünwald-Letnikov (G-L) fractional order differential definition instead of the derivative of the Gaussian function. Furthermore, a new edge detection mask based on the G-L fractional order calculus is suggested. Then, the quantitative relation curves between edge detection accuracy and the tuning parameters ( and ) are presented, which are helpful when selecting the optimal parameters. In addition, three effective evaluation criterions are introduced to assess the proposed method performance. We design a new edge detection mask based on the G-L definition, which is flexible in choosing the degree of fractional order derivative and capable of adjusting the direction of difference, and thus, can have extensive applications. We obtain the quantitative relation curves between edge detection accuracy and the parameters ( and ), which can be a guide in obtaining optimal parameters to extract desirable edges. We take many experiments using synthesize and real images, and compare the proposed method with five traditional edge detecting methods and three kinds of methods based on fraction calculus. The detection accuracy, detection efficiency and the robustness of the new method proposed in this paper are improved. In a digital image, a high-frequency component corresponds to edges and noises, an intermediate-frequency component corresponds to the texture detail, and a low-frequency component corresponds to the smoothing area. Using a fractional order differential for edge detection can completely extract weak edges and texture detail. Moreover, using a fractional order differential with suitable parameters can improve noise immunity capability. Edge-detecting efficiency is improved compared with other methods. Thus, the proposed method can be used in many real-time image-processing systems. Considerable experiment results show that the proposed method is a valid edge detector for a textured image, and even exhibit an evident advantage over other techniques. The proposed method is a significant extension of the traditional Canny detector. Overall, the proposed method is a valid and effective edge detection technique.关键词：edge;edge detection;Grünwald-Letnikov fractional order;fraction order derivative;Canny operator;texture image;weak edge5178|305|11更新时间：2024-05-07
- 摘要：Infrared small target detection is an important branch of computer vision applications. It has extensive applications in the military and civil fields, such as in precision guidance, security monitoring, and medical imaging systems. Infrared image preprocessing includes noise reduction and background suppression, and the main factor that affects small target detection is background clutter interference. Therefore, infrared image preprocessing prioritizes background suppression algorithms. Among numerous background suppression algorithms, the Robinson guard filter is a widely used filter based on a guard band. This filter cannot only achieve excellent suppression effect, but can also integrally retain the edge and internal information of small targets. However, the filtering mechanism of the Robinson guard filter is more sensitive to noise and cannot effectively deal with small targets with varying sizes. Given the limitation of the Robinson guard filter, this study proposes a small target detection method based on infrared background suppression. First, the background of infrared images is suppressed via Tophat transformation of the mathematical morphology to decrease the influence of noise on followup processing and reduce the sensitivity of the filtering mechanism of the Robinson guard filter to noise. Second, the modified Robinson guard filter is adopted to further suppress the background and to enhance the small targets of the processed images. Then, the regions of interest are identified using an adaptive threshold method. On this basis of this method, Unger smoothing is performed to remove small noise points. Then, we apply the local signal-clutter ratio (SCR) to eliminate residual background clutter after considering the difference between the gray value of the small targets and the gray value of the background, Finally, a moving pipeline filter is applied based on the motion continuity of the infrared small targets to exclude false positives, and the locations of the true small targets are accurately confirmed. We use three groups of image sequences with different backgrounds as experimental objects. Furthermore, we provide filtering performance (e.g., GSNRG and BSF) and detection performance (e.g., detection rate and false alarm rate) as evaluation indices. Test results show that the proposed method outperforms the Tophat algorithm, the max-mean algorithm, the Robinson guard filter, and the improved local contrast measure (ILCM) algorithm in terms of background suppression and detection performance. Compared with the traditional Robinson guard filter, the proposed method cannot only retain the characteristics of the small targets significantly, but can also improve the detection rates of the three groups of image sequences by 1.1%, 2%, and 11%, respectively. Moreover, it can decrease their false alarm rates by 14%, 12%, and 16%, respectively. The proposed method can achieve superior detection performance and can fully satisfy real-time requirements based on the test results. The experiment results demonstrate that the proposed method exhibits better performance in terms of background suppression and small target detection than the other four algorithms. The method also exhibits good adaptability in low SCR images. Meanwhile, it possesses high capability for real-time processing, which is conducive to realizing technological applications.关键词：infrared target detection;background suppression;Robinson guard filter;Unger smooth;pipeline filter3257|315|9更新时间：2024-05-07
- 摘要：Feature matching is one of the most important research topics in the field of image processing. However, most available methods fail to achieve satisfying quantitative and qualitative matches simultaneously. In this study, we introduced epipolar constraint into speeded-up robust features (SURF) feature matching, thereby achieving significant improvement. In this method, the SURF algorithm was adopted to detect the feature points of each studied image. Then, the fundamental matrix was calculated using random sample consensus (RANSAC) and was used to obtain the epipolars of all the points. Finally, a constraint was introduced into the epipolars to filter error matches. Consequently, significantly improved matches with enhanced quantity and quality were achieved. The experimental results indicate that compared with the old method, our method cannot only obtain matches with high accuracy but can further achieve an increase of twofold to eightfold in quantity. The process and implementation of the proposed method are simple and accurate. Moreover, the method can increase the number of correct matches and handle different types of images.关键词：image processing;feature matching;epipolar constraint;fundamental matrix;speeded-up robust features(SURF)3242|249|3更新时间：2024-05-07
- 摘要：Traditional visual tracking methods only involve the characteristic information extracted from the targets, and thus, the dense context of the targets are ignored. When extracting target feature information is difficult, tracking failure can easily result. To solve the aforementioned problem, an anti-occlusion visual tracking algorithm is presented based on spatio-temporal context learning. First, the spatio-temporal relationship between the target and local contexts is used to understand the spatio-temporal context model. Second, the confidence map is calculated using the prior context model and the learned spatio-temporal context model. Finally, the spatio-temporal context region is divided into blocks to consider occlusion discrimination. If the probability of occlusion is less than the set threshold, then the object location will be determined by maximizing the probability of the new confidence map. If the probability of occlusion exceeds the set threshold, then the target will be occluded. Object location and motion trajectory are estimated by matching sub-blocks and by filtering particles to realize anti-occlusion tracking in different levels. Conducting experiments in the benchmark data set demonstrates that the success rate of the STC-PF (Spatio-Temporal Context-Particle Filter)tracking method increases to over 80%. The center error is less than those of STC algorithms. With regard to the precondition of improving anti-occlusion capacity, the running speed of the STC-PF algorithm is higher than those of current popular algorithms and has minimal difference with that of the STC algorithm. The STC-PF algorithm can be applied to visual target tracking in complex conditions, such as illumination changes, object rotation, occlusion, and so on, with a certain instantaneity and efficiency. In particular, it exhibits good anti-occlusion capability and fast running speed in the case of a partially occluded target.关键词：visual tracking;spatio-temporal context;occlusion discrimination;sub-blocks matching;particle filter;confidence map4054|424|10更新时间：2024-05-07
- 摘要：Object tracking is a key issue in computer vision. A tracking algorithm based on compressive sensing theory has a high success rate. However, the efficiency of this algorithm requires further improvement. In addition, this algorithm has to deal with target occlusion. A tracking algorithm based on compressed sensing and particle swarm optimization (PSO) was proposed by focusing on the aforementioned issues. To improve the efficiency of a tracking algorithm based on compressed sensing, the PSO algorithm is incorporated into compression tracking. Furthermore, PSO was chosen over the method that required every other pixel to select target candidates. When a target is occluded, the proposed tracking algorithm can search the total image using PSO. The global search capability of PSO can be efficiently used by the proposed algorithm. This feature can significantly reduce the time required to find the target while improving the anti-occlusion capability of the tracking algorithm based on compressed sensing. The proposed algorithm is implemented on 20 publicly available challenging video sequences, and its performance is evaluated through a comparison with 7 state-of-art methods. The time-consuming process of tracking each frame, the average success rate, and the average deviation of the center position are obtained from the experiment. Experimental results on the 20 video frames show that the proposed algorithm significantly improves tracking efficiency and can adapt to both appearance changes and occlusion. Thus, the tracking success rate is significantly improved. The experimental data indicated that the average success rate reached 65.2 percent at an average of 155.5 frames per second, with an average center position deviation of 33.4. The tracking success rate of the proposed tracking algorithm reached over 85 percent in 9 video sequences, and the center position deviation reached 16 pixels or less in 11 video sequences. Compared with similar algorithms, the average success rate, average center position deviation, and average frames per second of the proposed tracking algorithm are relatively superior. In the compressive tracking algorithm that uses PSO to optimize the calculation of the number of search targets, the calculation count of target classification and recognition is reduced from the original 1 750 times to 120 times. The efficiency of the algorithm is significantly improved. Furthermore, when the target is occluded, the proposed algorithm uses PSO to search the entire image until the target reappears; once the target reappears, the proposed tracking algorithm resumes partial target search. Thus, the search capability of the tracking algorithm when the target is covered is improved. The proposed algorithm can find the reproduction target rapidly and accurately.关键词：computer vision;object tracking;compressive sensing;particle swarm optimization;feature extraction;naive Bayesian2396|296|0更新时间：2024-05-07
- 摘要：Building individual 3D dental models for patients is critical for the accurate localization and quantitative evaluation of computer dental restoration. A novel 3D tooth reconstruction method based on level set active contour model is proposed for the characteristics of teeth morphologic change and permutation of oral CT image sequences. Different models are adopted for different parts of tooth slices in the mapping mechanism between layers. A single hybrid level set model combined with the prior shape constraint energy, the edge gradient energy based on flux model, and the local region energy based on prior intensity is used to segment tooth contour in dental root slices; while a double hybrid level set model combined with regional competitive restraint energy is used for dental crown slices. These segmented contours are then applied to reconstruct 3D dental models. After testing numerous CT tooth slices at different positions, the results demonstrate that the proposed method provides better segmentation outcome and higher percentage accuracy than existing methods. The mean similarity index can reach 96%. The proposed hybrid level set model can efficiently overcome the interference of pulp cavity and alveolar bone as well as the problem of uneven intensity. Furthermore, the method can create a relatively accurate reconstruction of the 3D models for each individual tooth, which can lay a solid foundation for dental restoration planning and biomechanical analysis.关键词：tooth segmentation;CT image;level set;3D reconstruction;prior shape;prior intensity;gradient constraint3045|272|2更新时间：2024-05-07
- 摘要：Anomaly detector has become increasingly important in remote sensing data analysis and has been used in many applications, such as environmental and agricultural monitoring, geological exploration, and national defense security. According to special spectral content, an anomaly target has an obvious edge feature, which corresponds to a high frequency. By contrast, the background corresponds to a low frequency because of its smooth spectral content. Considering different spectral contents from the background, the anomaly target can be filtered out from the high frequency of the edge. A fast anomaly detector has been proposed to detect anomaly by linear filter of the spatial domain. However, texture and detail of clutter background also have the characteristic of high frequency. Linear filter has difficulty separating the anomaly from the clutter background accurately. Compared with bright background object, spatial salience of anomaly will be decreased. Furthermore, small size of anomaly will lead to subpixel anomaly, which will blur the edge feature of the target. A small anomaly target may not be successfully detected by a spatial filter. Conversely, cross analysis of a binary image reduced the complexity of computation. However, self-correlation of the large anomaly target will lead to a hollow effect in the center area. Inspired by the nonlinear filter mechanism of biotical vision, a bionic anomaly detection algorithm is proposed. In the natural world, a biotical vision system can accurately detect a small moving target, even in a cluttered environment. Redundancy information of the background will be inhibited because of its invariance on the spatial or temporal domain. Only features can be maintained as a high-order feature caused by a variance on the spatial and temporal domains. In fact, an anomalous spectral content of the target not only reflects a single band (spatial domain) but also reflects all the bands. Inspired by biotical vision, a correlated-type nonlinear filter is proposed to extract the high-order feature within the joint spatial and spectral domain. Like a moving target, the anomaly can be detected because of its spatial spectral wave, which contains spectral content of all bands. Simultaneously, the clutter background will be inhibited effectively because of its correlation with the spectral wave within the local spatial domain. Furthermore, the inner window is applied as a protective band, which can prevent the correlation of the anomaly target self, to avoid the hollow effect of a large anomaly target. Simulated and real data were applied to verify the utility of the proposed method. Experimental results show that the proposed anomaly detector has a good performance for small anomalies, which are rounded by clutter background. For a larger anomaly target, the hollow effect had to be removed within cross analysis by the protective band. This study proposed a bionic anomaly detector based on nonlinear filter. The high-order feature is extracted by nonlinear filter, with joint spatial and spectral information. The high-order feature has a strong robustness under the clutter background, particularly for a small target. Simultaneously, the inner window as protective band improves the hollow effect of a large anomaly target.关键词：remote sensing image;anomaly detection;clutter background;bionic algorithm;non-linear filter;false alarm rate2319|202|0更新时间：2024-05-07
- 摘要：To overcome nonconformity between the boundaries of a region using a segmentation algorithm and real object boundary, a segmentation optimization algorithm based on the fusion information of boundaries is presented in this study using the homogeneity characteristic of the objects and the salient boundary information in a high-resolution remote-sensing image. First, the canny edge filter was applied to extract edge information for remote-sensing images, and then the edge discontinuity problem was addressed to obtain the closed boundary. Second, the boundary and the initial segmentation results were fused, and new segmentation results were obtained. Finally, the new segmentation results under the closed boundary constraint were merged to obtain the optimization segmentation results based on the gray similarity criterion. This study presents a segmentation algorithm used to optimize the segmentation results obtained via MeanShift and eCognition. Furthermore, the rightly-segmented ratio (RR) is increased by 4% compared with the initial results, thereby verifying the effectiveness of the optimized algorithm. The optimized algorithm has extensive applicability and can optimize methods based on regions, boundaries, and clustering. This algorithm cannot only maintain the regional integrity of the object in high-resolution remote sensing, but can also keep the details of the object edge, thereby improving segmentation accuracy.关键词：high resolution imagery;image segmentation;edge feature;fusion;region merging3221|985|3更新时间：2024-05-07
- 摘要：Remote sensing image fusion based on sparse representation has achieved dramatic results. However, classical sparse representation faces two problems. First, the computational complexity of dictionary training is extremely high. Second, the classical sparse representation does not consider the similarity between image patches, which causes performance degradation in solving sparse coefficients. A new structural group sparse representation (SGSR), which has been successfully applied to image restoration and super-resolution image reconstruction, can effectively solve the two problems. To improve the accuracy and speed of sparse representation of remote sensing image fusion, this study presents a method of remote sensing image fusion based on SGSR. Considering that the spectral range of panchromatic image does not exactly override all the bands of a multispectral image, a luminance component derived from the adaptive weight coefficient of each band image is initially defined to reduce the spectral distortion. Then, the Euclidean distance between the gray scales of image patches is calculated. The most similar structure patches are selected to constitute the structural groups for luminance component image and panchromatic image, and these structural groups are regarded as the basic unit of dictionary learning and sparse representation. The adaptive dictionary learning method by singular value decomposition obtains the adaptive dictionary for every structural group. Furthermore, group sparse coefficient of the luminance component image and panchromatic image are calculated via group sparse representation algorithm. Finally, using the absolute maximum fusion rule, new sparse coefficients can be obtained by part of panchromatic image sparse coefficient substitution, and the high spatial resolution intensity image is reconstructed using the panchromatic image group dictionary and the new sparse coefficients. The high-resolution multispectral image is obtained via the general component substitution framework. SGSR is evaluated by the degraded and nondegraded image fusion experiments on Gaofeng-1, Quickbird, and Worldview images. In the degraded remote sensing image fusion experiment, five quantitative evaluation indices (CC, RMSE, ERGAS, UIQI, and SAM) are used to compare SGSR with the traditional methods (GIHS, PCA, and HPF), AWLP, and classical sparse representation (SR) method. Results show that five quantitative indices of SGSR are better compared with the other methods. In the nondegraded remote sensing image fusion experiment, psnr, grad, entropy, Dλ, Ds, and QNR are used to compare these image fusion methods. SGSR generally obtained better results, except that spectral information preservation performance of SGSR in the green plant region is slightly worse than those of AWLP. By comparing the speed of training dictionary, SGSR is approximately 10 times faster than SR. SGSR can improve spatial resolution and preserve spectral information in a remote sensing image and has better performance than the traditional methods, AWLP, and SR. Moreover, by comparing SGSR with SR, SGSR significantly reduces the computational complexity. Therefore, SGSR is suitable for multispectral image and panchromatic image fusion of all types of remote sensing images and may be applied in this field.关键词：remote sensing image fusion;adaptive dictionary group;structural group representation;sparse representation;general component substitution3003|364|7更新时间：2024-05-07
- Postal code：100190
- Tel：010-58887035/58887030/58887418 Email：jig@aircas.ac.cn
- Technical support is provided by Beijing Founder electronics co., LTD 京ICP备05080539号-4
京公网安备11010802024621 - It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
- Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
0