Search

This paper presents a new algorithm for fast exemplar-based image inpainting by using image pyramid and a novel pruning scheme. The proposed technique separates the inpainting process into three major steps. First the model determines which patch belonging to the inpainting region should be filled first. Then the pruning system analyzes the selected patch and weeds out regions having no chance to be the best matching patch. Finally, it performs a fast texture synthesis based on greedy search method in the remaining regions to fill the selected patch. Compared with previous methods, the proposed algorithm is much more efficient computationally. Experiments on synthetic and natural images show the advantages of the proposed algorithm, where it is shown to compare favorably with contemporary state-of-the-art schemes. [ABSTRACT FROM AUTHOR]

In this paper, a new approach has been proposed for improved facial expression recognition. The new approach is inspired by the compressive sensing theory and multi-resolution approach to facial expression problems. Initially, each image sample is decomposed into desired levels of its pyramids at different sizes and resolutions. At each level of the pyramid, features are extracted using a measurement matrix based on compressive sensing theory. These measurements are concatenated together to form a feature vector for the original image. The results obtained from the approach using three distance measurement classifiers (Manhattan, Euclidean, Cosine) and support vector machine are impressive and outperforms most of its counterpart algorithms in the literature using the same databases and settings. [ABSTRACT FROM AUTHOR]

In this paper, a new approach has been proposed for improved facial expression recognition. The new approach is inspired by the compressive sensing theory and multiresolution approach to facial expression problems. Initially, each image sample is decomposed into desired pyramid levels at different sizes and resolutions. Pyramid features at all levels are concatenated to form a pyramid feature vector. The vectors are further reinforced and reduced in dimension using a measurement matrix based on compressive sensing theory. For classification, a multilevel classification approach based on single-branch decision tree has been proposed. The proposed multilevel classification approach trains a number of binary support vector machines equal to the number of classes in the datasets. Class of test data is evaluated through the nodes of the tree from the root to its apex. The results obtained from the approach are impressive and outperform most of its counterparts in the literature under the same databases and settings. [ABSTRACT FROM AUTHOR]

In order to solve the ill-posed problem and restore image details， this paper proposed a two-level deconvolution based image blind restoration algorithm for wide field fluorescence microscopic images. Using both two-level deconvolution scheme and image pyramid structure，the proposed algorithm estimated latent images from coarse to fine. To suppress ill-posed problem，outer-level deconvolution applied total variation regularization term to both latent image and optical transfer function. Inner-level deconvoluton used residual image to restore details information further. Experiment results show that the proposed algorithm can recover details of wide field microscopic images with both artifacts and noises suppressed. Compared with other image blind restoration algorithm in recent years，the proposed algorithm takes less time to restore latent image. The estimated latent images not only have better visual quality but also have higher PSNR and image entropy. [ABSTRACT FROM AUTHOR]

This paper details a DSM (digital surface model) generation method using ZY-3 images based on object semi-global optimization. This method avoids the limitations imposed when building a match cost cube in the standard method. The proposed method combines semi-global optimization and an image pyramid to dynamically determine the search space of every pixel in the next pyramid layer according to the match result of the previous layer. Combining outlier detection technology and using mutual information and CENSUS as cost function, it realizes high-precision DSM generation from ZY3 images. We analyzed the key factors affecting DSM accuracy through experiments. [ABSTRACT FROM AUTHOR]

Currently, generating spatial image pyramid data mainly relies on down-sampling filtration, but so far, there is not any way to evaluate the effect of down-sampling. Herein, down-sampling and up-sampling were combined to form a pair of re-sampling filter, called RSFP, serving as an approximation of the current layer of pyramid data, can be used to evaluate the effect of the down-sampling filter. Based on RSFP, a pyramid-generating approach was built up in here, called it as TDFA. Its filtering depends on the texture direction of the pyramid image data. TDFA down-sampling PSNR was higher than the nearest neighbor interpolation, about 4.51-5.70 dB, while the latter was ever known the best down-sampling filter. The traditional JPEG compression method ignored the close correlations among the pyramid interlayer data, its compression process does not depend on the image texture features, which is not conducive to improving the compression ratio. The proposed RSFP and texture filtering method TDFA can effectively remove both correlations of pyramid image data, which respectively are the inter-layer correlations and its intra-layer texture correlations. Based on the re-sampling filter pair RSFP, combing these two means, an inter-layer compress method ICRS was created in this paper. Under the same reconstruction conditions, ICRS was found to increase the compression ratio 6.02 to 19.70 higher than the conventional AVS's I-frame algorithm, and on the whole, its compression ratio is 3 times or more high than that of JPEG algorithm, and there still is considerable room for improvement. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Image & Graphics is the property of Editorial Office of Journal of Image & Graphics and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Harris corner points were adopted as tracking objects， and scale space was introduced into corner point detection in order to extract Harris corner points in feature scale. Then curvature was computed to filter out false corners and enhance adaptability to scale change. Optical flow method was adopted for the tracking algorithm based on image pyramid， in which the optical flow iteratively was computed. And the tracking algorithm based on the optical flow error was proposed. That is， the trajectory error in the same frame with different time flow was used to evaluate the tracking situation. In this way， tracking failure was avoided when the tracking object is hidden， disappears or textural features change. Experimental results of different video sequences show that the proposed optical flow tracking algorithm based on improved corner extraction and image pyramid has better tracking performances. The feature points could be filtered effectively that lead to tracking failure with the introduction of optical flow error method， and the object positions are estimated accurately. [ABSTRACT FROM AUTHOR]

In this paper, a novel strategy for correcting both geometric and image intensity distortions of echo-planar imaging (EPI) MRI data is presented. To deal with small local distortions caused by rapid changes of the magnetic field, an improved multimodal registration framework using normalized mutual information (NMI) in combination with a multi-scale technique is presented to estimate a dense displacement field. To ensure the robustness of this high dimensional ill-posed inverse problem, a novel anisotropic regularization functional is used. In order to quantify geometric distortions, a new quality measure, called standardized contour distance (SCD), is introduced. It uses the outer structure shape (OSS) information as basis for the evaluation. The new registration method was evaluated with one monomodal phantom data set and two multimodal human brain data sets (BrainSuite trainings data, SPM Subject data). By comparing with recent and efficient techniques of the state of the art, in the monomodal case, the new approach achieves results comparable to the sum of squared differences as data term. In the multimodal cases, our new registration strategy improves the mean of the SCD from 0.96 ± 0.11 to 0.60 ± 0.13 in case of the SPM Subject data and from 0.92 ± 0.07 to 0.78 ± 0.11 in case of the BrainSuite trainings data. [ABSTRACT FROM AUTHOR]

Template matching method is one of the most significant object recognition techniques and it has many applications in the field of digital signal processing and image processing and it is the base for object tracking in computer vision field. The traditional template matching by correlation is performed between gray template image w and the candidate gray image f where the template's position is to be determined in the candidate image. This task can be achieved by measuring the similarity between the template image and the candidate image to identify and localize the existence of object instances within an image. When applying this method to colored image, the image must be converted to a gray one or decomposed to its RGB components to be processed separately. The current paper aims to apply the template matching technique to colored images via generating the quaternion Fourier transforms of both the template and candidate colored image and hence performing the cross-correlation between those transforms. Moreover, this approach is improved by representing both the image and template as pyramid multi-resolution format to reduce the time of processing. The proposed algorithm is implemented and applied to different images and templates using Matlab functions. [ABSTRACT FROM AUTHOR]

A novel exemplar based inpainting approach running on multi-scaled space is presented. The presented approach focuses on designing a novel processing architecture that works on a multi-resolution hierarchy. An inpainting process is designed to run recursively on the Laplacian image pyramid from coarse to fine level. Through the recursive process running on the image pyramid, both of structural features and texture informations near the target regions are permeated in the restored region. It reduces the artifact effects caused by exhaustive patch searching of conventional methods which work on a single layer. More plausible inpainting results and improvement on the processing speed are achieved when the presented method is applied in the experiments. [ABSTRACT FROM AUTHOR]

Real--time detection of the aerator state Is an important job in aquiculture. We propose a method to detect the state of the aerator based on corner optical flow and support vector machine (SVM) model. We firstly collect the videos of stopping and running states of the aerator through camera. Then, we extract two adjacent frames in sequence of the video frames, use the Harris algorithm to detect the interest points of spray in the former frame and calculate its optical low by the pyramid Lucaskanade algorithm according to the latter frame, thus the average displacement of the interest points between the two frames is obtained. Thirdly, we train the SVM model by the average dispiacement data of the interest points between the two frames at the earning phase, and utilize the trained SVM model to predict the status of the aerator accordingly at the detecting phase. In add-on, we introduce a method to eliminate the frames with abnormal average dispiacement of the interest points to improve the detection accuracy. Experimental results show that our method is robust and can be adapted to momtor the state of the aerator in real time under condltions such as various illumination, different shooting angles and distances with higher detection accuracy than that of the histogram thresholding method. [ABSTRACT FROM AUTHOR]

For detecting pedestrians in video and image fast and accurately, this paper proposed a hierarchical method for pedestrian detection, which could combine holistic information and local information. The method was based on random forests and used image pyramid for multi-layer information fusion. Firstly, it trained a holistic random forest classifier with dominant orientation templates (DOT) at the first low spatial resolution layer. And it could be used for detecting candidate areas for pedestrian. Secondly, it extracted image patches with offset vectors to learn the appearance model and geometric constraint with part-based random forest at the second high spatial resolution layer. Finally, it detected pedestrian accurately in candidate areas at the second layer by Hough voting. According to the theory analysis and experimental results, the method obtains lower computation complexity and higher precisions than previous works. Multi-layer information fusion can effectively solve the problem of fast and accurate pedestrian detection. [ABSTRACT FROM AUTHOR]

Salient region detection is important for many high-level computer vision tasks. The majority of previous works exploit element contrast to detect image saliency region. In this paper, we propose a novel approach to analyze saliency cues from multiple scales of image structure, using a multi-scale image abstraction. In each image layer global color contrast cue and color spatial distribution cue are integrated to generate a single-layer saliency map, and then the final saliency map can be obtained by across-scale adding several single-layer saliency maps. The proposed saliency estimation method abstracts unnecessary image detail, obtaining high quality saliency detection results. We have evaluated the results of our method on the two publicly available datasets MSRA-1000 and SED. The experimental results on these datasets demonstrate the effectiveness of the approaches against the other approaches to analyze image saliency. [ABSTRACT FROM AUTHOR]

Infrared texture is an important feature in identifying scenery. To simulate infrared image texture effectively at different distances, we propose a model of infrared image texture generation based on scenery space frequency and the image pyramid degradation principle. First, we build a spatial frequency filter model based on imaging distance, taking into account the detector's maximum spatial frequency, and use the filter to process a "zero" distance infrared image texture. Second, taking into consideration the actual temperature difference of the scenery's details due to variation of the imaging distance and the effect of atmospheric transmission, we compare the actual temperature difference with the minimum resolvable temperature difference of the thermal imaging system at a specific frequency and produce a new image texture. The results show that the simulated multiresolution infrared image textures produced by the proposed model are very similar (lowest mean square error = 0.51 and highest peak signal-to-noise ratio = 117.59) to the images captured by the thermal imager. Therefore, the proposed model can effectively simulate infrared image textures at different distances. [ABSTRACT FROM AUTHOR]

Traditional single-lens vertical photogrammetry can obtain object images from the air with rare lateral information of tall buildings. Multi-view airborne photogrammetry can get rich lateral texture of buildings, while the common area-based matching for oblique images may lose efficacy because of serious geometric distortion. A hierarchical dense matching algorithm is put forward here to match two oblique airborne images of different perspectives. Based on image hierarchical strategy and matching constraints, this algorithm delivers matching results from the upper layer of the pyramid to the below and implements per-pixel dense matching in the local Delaunay triangles between the original images. Experimental results show that the algorithm can effectively overcome the geometric distortion between different perspectives and achieve pixel-level dense matching entirely based on the image space. [ABSTRACT FROM AUTHOR]

High Performance Computing (HPC) systems are nowadays more and more heterogeneous. Different processor types can be found on a single node including accelerators such as Graphics Processing Units (GPUs). To cope with the challenge of programming such complex systems, this work presents a domain-specific approach to automatically generate code tailored to different processor types. Low-level CUDA and OpenCL code is generated from a high-level description of an algorithm specified in a Domain-Specific Language (DSL) instead of writing hand-tuned code for GPU accelerators. The DSL is part of the Heterogeneous Image Processing Acceleration ( HIPA cc ) framework and was extended in this work to handle grid hierarchies in order to model different cycle types. Language constructs are introduced to process and represent data at different resolutions. This allows to describe image processing algorithms that work on image pyramids as well as multigrid methods in the stencil domain. By decoupling the algorithm from its schedule, the proposed approach allows to generate efficient stencil code implementations. Our results show that similar performance compared to hand-tuned codes can be achieved. [ABSTRACT FROM AUTHOR]

Poisson composition and gradient domain are typical and traditional composition technologies. Their main function is to create a seamless composited image. However, when the original image and the target image have quite a few differences with regard to their features (color, sharpness, noise, texture and so on), the composited image is unrealistic. We use some innovative methods to deal with the problem and there are three main stages considered in the paper. The first stage is to deal with the original image's color ahead of schedule. This is to make the original image as similar as the target image and it contributes to get a realistic composited image. The second stage is to achieve the image's multi-scale composition with wavelet pyramid. The third stage is that we use BLF filter and an image pyramid to preserve the composited image's detail after the image composition. Image composition's optimization is based on the three stages. [ABSTRACT FROM AUTHOR]

The feature points of the moving target in the video can be tracked through optical flow algorithm. When the target exists a movement with a relatively large scale, it is difficult to meet the image consistency hypothesis of optical flow, which results in the loss of tracked feature points. Concerning this problem, a method of moving human feature points tracking based on Lucas-Kanade pyramidal optical flow algorithm was proposed. First, the moving region of the human was obtained by the difference between the consecutive frames. Then, some feature points of the start frame were detected with the SIFT algorithm. Finally, the feature points were tracked in the subsequent frames through the image pyramidal optical flow. The experimental results suggest that the algorithm performs well on the feature points tracking of large scale movement and the tracking accuracy is improved. [ABSTRACT FROM AUTHOR]

Abstract: We present an effective traversal algorithm and a hardware architecture to accelerate inverse displacement mapping. This includes a set of techniques that are used to reduce the number of iterative steps that are performed during inverse displacement mapping. For this purpose, we present two algorithms to reduce the number of descending steps and two algorithms to improve the ascending process. All these techniques are combined; we observe up to 66% reduction in the number of iterative steps as compared to other pyramidal displacement-mapping algorithms. We also propose a novel displacement-mapping hardware architecture based on the novel techniques. The experimental results obtained from the FPGA and ASIC evaluation demonstrate that our novel architecture offers many benefits in terms of chip area, power consumption, and off-chip memory accesses for mobile GPUs. [Copyright &y& Elsevier]

• An out-of-plane modal properties extraction method was developed. • Stereophotogrammetry was used to obtain out-of-plane responses. • Continuous mode shapes can be reconstructed from multi-level image pyramid. • The physics-based graphics model was employed to explore the proposed method. • The proposed method was experimentally verified by a three-story model building. Understanding the dynamic behavior of as-built structures is important, because structural condition can often be inferred from changes in its dynamic response. To this end, modal properties such as natural frequencies, damping ratios, and mode shapes can be extracted from acceleration measurements. Associated installations require sensors, cabling, and data acquisition systems that can be expensive and time consuming. Alternatively, computer vision-based approaches have been proposed that offer non-contact measurements, as well as substantial cost savings. However, such approaches usually focus only on the structural component responses that are in the plane of the image; researchers seldom consider the structural response that is out-of-plane (i.e., perpendicular to the image plane). For complex structural component that possess asymmetries, the out-of-plane behavior is critical to understanding their dynamic response. Determination of dynamic response of the structure requires that multiple cameras are used and can be computationally demanding. In this study, imagery from a commercially available stereo camera is combined with a multi-level image pyramid approach and operational modal analysis (OMA) to extract out-of-plane modal properties. The measurements are first padded to reduce the distortion effect caused by the image pyramid. The responses are then compressed and decomposed into sub-bands using an image pyramid decomposition; followed by extraction of the modal properties using the OMA approach known as frequency-domain stochastic subspace identification. The advantages and limitations of the proposed approach are illustrated numerically using a physics-based graphics model (PBGM) of a continuous beam. Subsequently, experimental validation is conducted for a 3-story model building using the Intel® RealSenseTM D415 depth camera. Modal properties are shown to be determined quickly, with high accuracy and noise robustness. Moreover, detailed near-continuous mode shapes are obtained using multi-level image pyramid reconstruction. These results demonstrate that the proposed approach can give an accurate picture of the dynamic characteristics of a structure, offering the potential for effective long-term structural health monitoring for important structural component. [ABSTRACT FROM AUTHOR]

Recently, the local binary patterns (LBP) have been widely used in the texture classification. The LBP methods obtain the binary pattern by comparing the gray scales of pixels on a small circular region with the gray scale of their central pixel. The conventional LBP methods only describe microstructures of texture images, such as edges, corners, spots and so on, although many of them show good performances on the texture classification. This situation still could not be changed, even though the multi-resolution analysis technique is adopted by LBP methods. Moreover, the circular sampling region limits the ability of the conventional LBP methods in describing anisotropic features. In this paper, we change the shape of sampling region and get an extended LBP operator. And a multi-structure local binary pattern (Ms-LBP) operator is achieved by executing the extended LBP operator on different layers of an image pyramid. Thus, the proposed method is simple yet efficient to describe four types of structures: isotropic microstructure, isotropic macrostructure, anisotropic microstructure and anisotropic macrostructure. We demonstrate the performance of our method on two public texture databases: the Outex and the CUReT. The experimental results show the advantages of the proposed method. [ABSTRACT FROM AUTHOR]

We present a method to create high-quality sampling filters by combining a prescribed number of texels from several resolutions in a mipmap. Our technique provides fine control over the number of texels we read per texture sample so that we can scale quality to match a memory bandwidth budget. Our method also has a fixed cost regardless of the filter we approximate, which makes it feasible to approximate higher-quality filters such as a Lánczos 2 filter in real-time rendering. To find the best set of texels to represent a given sampling filter and what weights to assign those texels, we perform a cardinality-constrained least-squares optimization of the most likely candidate solutions and encode the results of the optimization in a small table that is easily stored on the GPU. We present results that show we accurately reproduce filters using few texel reads and that both quality and speed scale smoothly with available bandwidth. When using four or more texels per sample, our image quality exceeds that of trilinear interpolation. [ABSTRACT FROM AUTHOR]