Showing papers on "Subpixel rendering published in 2000"

Systematic errors in digital image correlation caused by intensity interpolation

Abstract: Recently, digital image correlation as a tool for surface defor- mation measurements has found widespread use and acceptance in the field of experimental mechanics. The method is known to reconstruct displacements with subpixel accuracy that depends on various factors such as image quality, noise, and the correlation algorithm chosen. How- ever, the systematic errors of the method have not been studied in detail. We address the systematic errors of the iterative spatial domain cross- correlation algorithm caused by gray-value interpolation. We investigate the position-dependent bias in a numerical study and show that it can lead to apparent strains of the order of 40% of the actual strain level. Furthermore, we present methods to reduce this bias to acceptable lev- els. © 2000 Society of Photo-Optical Instrumentation Engineers. (S0091-3286(00)00911-9)

Constrained subpixel target detection for remotely sensed imagery

Abstract: Target detection in remotely sensed images can be conducted spatially, spectrally or both. The difficulty of detecting targets in remotely sensed images with spatial image analysis arises from the fact that the ground sampling distance is generally larger than the size of targets of interest in which case targets are embedded in a single pixel and cannot be detected spatially. Under this circumstance target detection must be carried out at subpixel level and spectral analysis offers a valuable alternative. In this paper, the problem of subpixel spectral detection of targets in remote sensing images is considered, where two constrained target detection approaches are studied and compared. One is a target abundance-constrained approach, referred to as nonnegatively constrained least squares (NCLS) method. It is a constrained least squares spectral mixture analysis method which implements a nonnegativity constraint on the abundance fractions of targets of interest. Another is a target signature-constrained approach, called constrained energy minimization (CEM) method. It constrains the desired target signature with a specific gain while minimizing effects caused by other unknown signatures. A quantitative study is conducted to analyze the advantages and disadvantages of both methods. Some suggestions are further proposed to mitigate their disadvantages.

Robust image registration using log-polar transform

Abstract: This paper describes a hierarchical image registration algorithm for affine motion recovery. The algorithm estimates the affine transformation parameters necessary to register any two digital images misaligned due to rotation, scale, shear, and translation. The parameters are computed iteratively in a coarse-to-fine hierarchical framework using a variation of the Levenberg-Marquadt nonlinear least squares optimization method. This approach yields a robust solution that precisely registers images with subpixel accuracy. A log-polar registration module is introduced to accommodate arbitrary rotation angles and a wide range of scale changes. This serves to furnish a good initial estimate for the optimization-based affine registration stage. We demonstrate the hybrid algorithm on pairs of digital images subjected to large affine motion.

PIV Error Correction

Abstract: A non-post-interrogation method of reducing subpixel errors and eliminating spurious vectors from particle image velocimetry (PIV) results is presented. Unlike methods that rely on the accuracy or similarity of neighboring vectors, errors are eliminated before correlation information is discarded using available spatial and/or temporal data. Anomalies are removed from the data set through direct element-by-element comparison of the correlation tables calculated from adjacent regions. The result is a correction method that improves subpixel accuracy and effective spatial resolution and is highly robust to out-of-boundary particle motion, particle overlap, inter-particle correlations, and electronic and optical imaging noise.

Measuring earthquakes from optical satellite images

Abstract: Systeme pour l’Observation de la Terre images are used to map ground displacements induced by earthquakes. Deformations (offsets) induced by stereoscopic effect and roll, pitch, and yaw of satellite and detector artifacts are estimated and compensated. Images are then resampled in a cartographic projection with a low-bias interpolator. A subpixel correlator in the Fourier domain provides two-dimensional offset maps with independent measurements approximately every 160 m. Biases on offsets are compensated from calibration. High-frequency noise (0.125 m-1) is ∼0.01 pixels. Low-frequency noise (lower than 0.001 m-1) exceeds 0.2 pixels and is partially compensated from modeling. Applied to the Landers earthquake, measurements show the fault with an accuracy of a few tens of meters and yields displacement on the fault with an accuracy of better than 20 cm. Comparison with a model derived from geodetic data shows that offsets bring new insights into the faulting process.

Optimal filtering for patterned displays

Abstract: Displays with repeating patterns of colored subpixels gain spatial resolution by setting individual subpixels rather than by setting entire pixels. This paper describes optimal filtering that produces subpixel values from a high-resolution input image. The optimal filtering is based on an error metric inspired by psychophysical experiments. Minimizing the error metric yields a linear system of equations, which can be expressed as a set of filters. These filters provide the same quality of font display as standard anti-aliasing at a point size 25% smaller. This optimization forms the filter design framework for Microsoft's ClearType.

Generalized constrained energy minimization approach to subpixel target detection for multispectral imagery

Abstract: Subpixel detection in multispectral imagery presents a chal- lenging problem due to relatively low spatial and spectral resolution. We present a generalized constrained energy minimization (GCEM) ap- proach to detecting targets in multispectral imagery at subpixel level. GCEM is a hybrid technique that combines a constrained energy mini- mization (CEM) method developed for hyperspectral image classification with a dimensionality expansion (DE) approach resulting from a gener- alized orthogonal subspace projection (GOSP) developed for multispec- tral image classification. DE enables us to generate additional bands from original multispectral images nonlinearly so that CEM can be used for subpixel detection to extract targets embedded in multispectral im- ages. CEM has been successfully applied to hyperspectral target detec- tion and image classification. Its applicability to multispectral imagery is yet to be investigated. A potential limitation of CEM on multispectral imagery is the effectiveness of interference elimination due to the lack of sufficient dimensionality. DE is introduced to mitigate this problem by expanding the original data dimensionality. Experiments show that the proposed GCEM detects targets more effectively than GOSP and CEM without dimensionality expansion. © 2000 Society of Photo-Optical Instrumenta- tion Engineers. (S0091-3286(00)01205-8)

Comparative analysis of hyperspectral adaptive matched filter detectors

Abstract: Real-time detection and identification of military and civilian targets from airborne platforms using hyperspectral sensors is of great interest. Relative to multispectral sensing, hyperspectral sensing can increase the detectability of pixel and subpixel size targets by exploiting finer detail in the spectral signatures of targets and natural backgrounds. A multitude of adaptive detection algorithms for resolved or subpixel targets, with known or unknown spectral characterization, in a background with known or unknown statistics, theoretically justified or ad hoc, with low or high computational complexity, have appeared in the literature or have found their way into software packages and end-user systems. The purpose of this paper is threefold. First, we present a unified mathematical treatment of most adaptive matched filter detectors using common notation, and we state clearly the underlying theoretical assumptions. Whenever possible, we express existing ad hoc algorithms as computationally simpler versions of optimal methods. Second, we assess the computational complexity of the various algorithms. Finally, we present a comparative performance analysis of the basic algorithms using theoretically obtained performance characteristics. We focus on algorithms characterized by theoretically desirable properties, practically desired features, or implementation simplicity. Sufficient detail is provided for others to verify and expand this evaluation and framework. A primary goal is to identify best-of-class algorithms for detailed performance evaluation.

Maximum-likelihood template matching

Abstract: In image matching applications such as tracking and stereo matching, it is common to use the sum-of-squared-differences (SSD) measure to determine the best match for an image template. However, this measure is sensitive to outliers and is not robust to template variations. We describe a robust measure and efficient search strategy for template matching with a binary or greyscale template using a maximum-likelihood formulation. In addition to subpixel localization and uncertainty estimation, these techniques allow optimal feature selection based on minimizing the localization uncertainty. We examine the use of these techniques for object recognition, stereo matching, feature selection, and tracking.

Subpixel-precise extraction of lines and edges

Abstract: Novel approaches to extract curvilinear structures, i.e., lines and edges, from 2D images are proposed. For lines, explicit geometric models for line profiles are used to analyze their behavior in scale-space. From this analysis, algorithms to extract lines and their widths with subpixel resolution are derived, and it is shown that the line positions and widths are necessarily biased by the smoothing used in the extraction. Since the mapping that describes the bias is invertible, it can be removed, leading to unbiased and hence accurate results. To extract edges, they are regarded as bright lines in the gradient image. Furthermore, by a scale-space analysis it is shown why line and edge junctions often cannot be extracted. From this analysis, a new method to extract complete junction information is derived. Finally, the line and edge extraction algorithms are extended to multispectral images.

Subpixel motion estimation from sequences of video images

Abstract: We develop a new class of multi-image gradient-based algorithms, and a discrete Fourier transform based algorithm to detect subpixel motions of objects in video images. Because of their enormous practical importance, we make estimators of amplitude and phase of temporal sinusoidal motion using both methods. We show that to improve motion estimates of existing gradient-based algorithms, it is not suÆcient to improve spatial gradient estimates alone; it is necessary to improve both spatial and temporal gradient estimates. We use data in many images to estimate spatial and temporal derivatives to high accuracy. By using many images, we are also able to compensate for the blur caused by the nite image acquisition times. Through analysis of simple images and through simulations, we show that the inherent bias of multi-image gradient-based methods can be made arbitrarily small for small motions. However, for large motions, multi-image gradient based methods cease to perform well. We simulate the performance of our algorithms in the presence of noise typical of optical microscopes and scienti c grade cameras. These simulations show that the sinusoidal estimators we create achieve errors below 0.001 pixels and 0.001 radians for amplitudes smaller than 1.2 pixels. However, for motions larger than 2 pixels, the amplitude errors are larger than 0.1 pixels. We show that Fourier transform based methods are limited by bias inherent to the method for the analysis region sizes that interest us. In the presence of noise typical for optical microscopes, the sinusoidal estimator we create achieves 0.1 pixel accuracy and 0.01 radian accuracy. These inaccuracies are greater than those of already existing algorithms. We experimentally test the performance of the multi-image gradient-based sinusoidal estimators. We show that the algorithms achieve nanometer accuracy and precision for motions below 500 nanometers. The results agree well with the predicted performance of the algorithm based on simulations. We also show that the algorithms are consistent to within a nanometer across regions of the same moving object with very di erent brightness These features of the new algorithms represent important improvements over existing algorithms. Thesis Supervisor: Dennis M. Freeman Title: Associate Professor of Electrical Engineering Acknowledgments Dr. C. Quentin Davis was invaluable by providing comments and by participating in several insightful conversations. Special thanks to Professor John Wyatt who helped me remain excited about this project. He also deserves credit for introducing me to Trefethen's book [37] which gave me important insights into some of the problems I faced. Professor Dennis M. Freeman deserves recognition for reasons too numerous to list here. Denny deserves special credit for teaching me how to write technically. Proof readers helped me develop this work from a disorganized, unintelligible grouping of ideas into an understandable document. I owe a great deal of thanks to Dr. Werner Hemmert, Jekwan Ryu, Michael Gordon, Dr. C. Quentin Davis, and of course Professor Dennis. M. Freeman. I am grateful for the support of the Fannie and John Hertz Organization who pays the great majority of my costs as a student. This work was also supported by a grant from DARPA (F3060297-2-0106). Finally, I recognize some of the outstanding work in the literature. While H.H. Nagel has produced several overly-long papers, a few of them are brilliant and essential to the understanding of gradient-based methods. Ng and Solo's [33] introduction of Sprent's work [35] into the motion estimation eld is an essential ingredient to the noise analysis of gradient-based methods. Davis and Freeman's work [7, 8, 9, 10] is crucial to the subpixel estimation eld, but not because the work is so revolutionary. Davis and Freeman deserve respect because they had the audacity to think their methods would work. Finally, Horn's initial work in the eld was instrumental in the development of optical ow and gradient-based methods.

Super-resolving imaging system

Abstract: A super-resolving imaging apparatus employs diffractive optical elements placed on the imaging lens. This element, and the use of a modified Scheimpflug arrangement allow the conversion of degrees of freedom in one axis of a field of view to a larger degree of freedom in another axis in order to obtain a high resolution image with a wide depth of focus and large field of view. Replicas created by the diffractive elements are mutually shifted by subpixel amounts, and are combined using a Gabor transform, which is facilitated by a spatial mask placed over the detector array. The apparatus is suitable for performing distance estimation on an object within the field of view.

A parametric template method and its application to robust matching

Abstract: We propose a template-matching method in which a parametric template space is constructed from a given set of template images then quickly matched to a reference image. In this method geometrical changes in an image object due to translation, rotation or scaling, and non-geometrical changes such as illumination variations or individual variation between objects are represented by template images in the parametric template space. The method also provides a subpixel matching scheme that yields a high-precision estimation of the object position. Experiments using real images have confirmed the effectiveness of the method.

Improving image display quality by adaptive subpixel rendering

Abstract: A method for displaying an image comprises receiving pixel data representing an image, determining an image feature in the received pixel data, and generating a subpixel drive correction signal based on the determined image. The generated subpixel drive correction signal can be based on display attributes, and can be used to generate a pixel drive signal.

Pixel border for improved viewability of a display device

Abstract: A display device having a display matrix including a pixel border of width x and located around the edge locations of the matrix for improved viewability. In particular, the border can be several pixels wide, e.g., 1

Subpixel resolution binocular visual tracking using analog VLSI vision sensors

Abstract: This paper describes the application of an analog VLSI vision sensor to active binocular tracking. The sensor outputs are used to control the vergence angles of the two cameras and the tilt angle of the head so that the center pixels of the sensor arrays image the same point in the environment. One distinguishing feature of the sensor used here is the possibility to resolve target displacement with subpixel resolution via a phase-based algorithm, which integrates information over multiple pixels.

Special sensor masking for exceeding system geometrical resolving power

Abstract: The resolution of many viewing and staring systems is often restricted by the spatial limited resolution of its sensing device rather than by diffraction limits related to the optical system. This spatial resolution of the sensor, when limited by the pixels' dimensions, is coined hereby "geometrical resolution." We suggest a technique for overcoming this limit, thus obtaining "geometrical superresolution." The proposed approach is based on capturing a set of images, interlacing their pixels and applying special filtering over the interlaced image. The number of the captured images N corresponds to the desired resolution improvement. Each image in the set is captured after a lateral shift of the sensor by a subpixel distance of ?x/N, where ?x is the size of the sensor's pixel. The main contribution is in designing a special energetic efficient mask that is attached to the sensing plane. Without this mask, unrecoverable resolution loss prevents a qualitative reconstruction to be obtained.

Controllable pixel border for improved viewability of a display device

Abstract: A display device having a display matrix (m+2x by n+2x) including an active, e.g., controllable, pixel border located around the edge locations of a frame buffer matrix for improved character viewability. The border can be several pixels wide, e.g., 1

Antialiasing method and image processing apparatus using same

Abstract: Disclosed are an antialiasing method and an image processing apparatus using the same, capable of high-quality image display without any significant reduction in processing speed and without any significant increase in the apparatus scale. Pixel data contains information on a subpixel mask indicative of region which a polygon occupies within a pixel. Based on data sets consisting of the subpixel masks and color data contained in the pixel data, display colors are determined on a pixel-by-pixel basis.

Intelligent image correlation using genetic algorithms for measuring surface deformations in the autonomous inspection of structures

Abstract: Summary form only given. Presents an intelligent digital image correlation technique that uses genetic algorithms to estimate surface displacements and strains for autonomous inspection of structures. Speckle patterns are spray painted on the surface of interest and pictures taken before and during loading using off-the-shelf CCD quickcams. Subpixel accuracy, required for measuring displacements and strains accurately, is obtained by using interpolation methods. An innovative adaptive scheme is used to develop a cost surface which is used to match the before and after image subsets with the help of genetic algorithms to search for the optimal match. Some potential applications for this work are to extract real time information on surface displacements and strains on aircraft in flight, submarines under water and civil infrastructures such as bridges and buildings.

Map-based road detection in spaceborne synthetic aperture radar images based on curvilinear structure extraction

Abstract: This paper presents an automatic map-based road detection algorithm for spaceborne synthetic aperture radar (SAR) images. Our goal is to find roads in a SAR image with subpixel accuracy with the help of a digital map. There are location errors between the digital map and the geocoded SAR image, which are about 20 to 30 pixels, and we adopt a coarse-to-fine strategy to reduce it. In the coarse matching step, we roughly find the locations of roads by a simple search using water areas or a generalized Hough transform based on digital map information. The fine matching step detects roads accurately by using the active contour model (snake). The input of the snake operation is the potential field constructed from the extracted ridges or ravines of curvilinear structures in the SAR image. Experimental results show that our algorithm detects roads with average error of less than one pixel. © 2000 Society of Photo- Optical Instrumentation Engineers. (S0091-3286(00)01309-X)

Simultaneous multi-frame subpixel boundary definition using toboggan-based intelligent scissors for image and movie editing

Abstract: Intelligent Scissors is an interactive image segmentation tool that allows a user to select piece-wise globally optimal contour segments (based on an optimal path search in a graph) that correspond to a desired object boundary. This dissertation uses tobogganing to raise the granularity of the image primitive above the pixel level, producing a region-based basic processing unit that is object-centered rather than device-dependent. The resulting region-based elements form the basis for several contributions to the field of computer vision general and to Intelligent Scissors in particular. These contributions reduce the human time and effort needed for object selection with Intelligent Scissors while simultaneously increasing the accuracy of boundary definition. The region-based image primitives resulting from tobogganing form the basis for a graph formulation that is many times smaller than the pixel-based graph used previously by Intelligent Scissors, thus providing faster, more interactively responsive optimal path computations. The object-centered atomic units also provide an efficient and consistent framework in which to compute a 4-parameter edge model, allowing subpixel boundary localization, noise-independent edge blur adjustment, and automatic alpha matte generation and color separation of boundary transition pixels. The increased size of the basic processing unit also facilitates an edge confidence measure that forms the basis for two new techniques called confidence threshold snapping and live-wire path extension, which further reduce the human burden involved with object boundary definition by automatically finding and following object boundaries. Finally, this dissertation presents a new paradigm for simultaneously interacting with multiple frames from a temporal image sequence by parallelizing both the user input and the interactive visual feedback, thus allowing a user to interact with a montage of image frames in order to define the boundary of a moving object while adhering to the same interactive style that has demonstrated to be effective for the single-image Intelligent Scissors.

Compensating for aperture parallax distortion in tiled displays

Abstract: A tiled display may include subpixels that may be partially occluded by an overlaying matrix. The matrix hides the joints between adjacent tiles. The partial occlusion of one subpixel of a pixel may result in chromatic shifts and/or luminance reduction. The partially occluded subpixel may be compensated for by providing an extra light producing subpixel of the same color on the opposite side of the matrix opening.

Cloud-top height estimation from satellite stereopairs for weather forecasting and climate change analysis

Abstract: Within the EU project Cloudmap, cloud-top heights should be estimated using satellite stereo images at very high temporal resolution. Since such images are not provided by operational sensors, MOMS-2P and ATSR2 images of lower temporal resolution were used instead. The MOMS stereopair was oriented with subpixel accuracy using GCPs from 1:50,000 topographic maps and Kratky’s sensor model. Preprocessing for noise reduction, cloud stripes removal and contrast enhancement was applied. For the estimation of cloud-top heights, the images were resampled to 288 m and geometrically constrained least squares matching, using image pyramids and an interest operator, as well as varying parameters was used. The results were checked by visual inspection and comparison to semi-automatically measured points in the original resolution images. Automatic blunder detection using two tests were also applied. Matching led to large blunders in land areas between clouds or close to cloud boundaries. Excluding these blunders (error > 1100 m), matching showed an RMS of ca. 0.2 pixel, exhibiting a very high accuracy potential. A matching geometric transformation using rotations and radiometric equalization during the iterations showed slightly better results compared to the other matching versions. ATSR2 images were matched with a similar approach, however without geometric constraints, as the input images were rectified. Due to differences between the images which vary spatially, varying matching parameters are optimal for each image region. First steps in combining matching results from such varying matching versions have been performed. Both datasets showed similar matching problems due to surface discontinuities, mixing of surfaces than are neighbouring in image space but differ in height, and often large illumination differences (even with along-track stereo and small time acquisition differences).

A simple underwater video system for laser tracking

Abstract: This paper presents a video system for underwater range finding and target tracking. The method depends on geometrical triangulation using a video camera, a laser diode, and an object plane. This system, which is completely controlled by a personal computer, provides excellent capabilities of efficient calibration and accurate range data processing. Through the calculation of the intensity center of the laser spot, we refine the performance of measurement up to subpixel accuracy. This system employs polynomial interpolation to calibrate the measurement error resulting from geometrical distortion and lens aberrations. The results show that the system can produce excellent accuracy for range finding and target tracking both in air and underwater.

Super-resolution from noisy image sequences exploiting a 2D parametric motion model

Abstract: We propose a low cost scheme for reconstructing high resolution images from noisy, and eventually blurred image sequences. The super-resolution is achieved by an iterative back projection method. To account for noise in image sequence, we first apply a spatio-temporal Wiener filter computed via a 3D DFT. In the filtering process, we need to compensate for apparent motion to ensure proper results. Furthermore, the knowledge of subpixel motion is necessary for super-resolution. In both cases, we exploit a parametric motion model to keep a good trade-off between accuracy and computation time.

Morphological pyramid image registration

Abstract: We propose an intensity-based morphological pyramid image registration algorithm. This approach utilizes the global affine transformation model, also considering radiometric changes between images. With the morphological pyramid structure, Levenberg-Marquardt optimization, and bilinear interpolation, this algorithm can be implemented hierarchically and iteratively with capability of measuring, to subpixel accuracy, the displacement between images subjected to simultaneous translation, rotation, scaling, and shearing. The morphological pyramid shows better performance than the Gaussian pyramid in this matching technique.