Scene Parameters Research Articles

Exchanging Faces in Images

Abstract Pasting somebody's face into an existing image with traditional photo retouching and digital image processing tools has only been possible if both images show the face from the same viewpoint and with the same illumination. However, this is rarely the case for given pairs of images. We present a system that exchanges faces across large differences in viewpoint and illumination. It is based on an algorithm that estimates 3D shape and texture along with all relevant scene parameters, such as pose and lighting, from single images. Manual interaction is reduced to clicking on a set of about 7 feature points, and marking the hairline in the target image. The system can be used for image processing, virtual try‐on of hairstyles, and face recognition. By separating face identity from imaging conditions, our approach provides an abstract representation of images and a novel, high‐level tool for image manipulation. Categories and Subject Descriptors (according to ACM CCS): I.3.7 [Computer Graphics]: Animation

Distortion-tolerant 3-D object recognition by using single exposure on-axis digital holography

We present a distortion-tolerant 3-D object recognition system using single exposure on-axis digital holography. In contrast to distortion-tolerant 3-D object recognition employing conventional phase shifting scheme which requires multiple exposures, our proposed method requires only one single digital hologram to be synthesized and used for distortion-tolerant 3-D object recognition. A benefit of the single exposure based on-axis approach is enhanced practicality of digital holography for distortion-tolerant 3-D object recognition in terms of its simplicity and high tolerance to external scene parameters such as moving targets. This paper shows experimentally, that single exposure on-axis digital holography is capable of providing a distortion-tolerant 3-D object recognition capability.

Applications of Depth Metadata in a Production System

This paper discusses applications for the use of depth metadata acquired with a production system that is under development in the EU IST-MetaVision project. The aim of the project is to develop a camera and production system to capture, store, and distribute program material that meets the demands of both the film and television industries. The key idea for cost-effective processing is to acquire and store metadata in addition to essence data (image material) such as camera and scene parameters. Available depth-sensing techniques are reviewed in order to identify suitable methods. At the current stage of technology no single technique covers all practical production situations, so several complementary techniques and a framework to integrate them are proposed. The requirements for representation and sensing of depth information are discussed for specific applications, and initial results are presented. Applications include the creation of special effects in post-production, optimized image coding, and (interactive) stereo viewing 3-D television (3-DTV).

A canonical problem in electromagnetic backscattering from buildings

In this paper, a geometric and electromagnetic model of a typical element of urban structure is presented, in order to analytically evaluate in closed form its electromagnetic return to an active microwave sensor. This model can be used to understand what information on geometric and dielectric properties of a building can be extracted from microwave remote sensing data. The geometrical model consists of a rectangular parallelepiped whose vertical walls form a generic angle with respect to the sensor line of flight. The parallelepiped is placed on a rough surface. The radar return from such a structure can be decomposed into single-scattering contributions from the (rough) ground, the building roof (a plane surface in our model), and vertical walls and multiple scattering contributions from dihedral structures formed by vertical walls and ground. In our model, single-scattering contributions are evaluated by using either physical optics (PO) or geometrical optics (GO), depending on surface roughness. In order to account for multiple scattering between buildings and terrain, we use GO to evaluate the field reflected by the smooth wall toward the ground (first bounce) or the sensor (second or third bounce) and GO or PO (according to ground surface roughness) to evaluate the field scattered by the ground toward the wall (first or second bounce) or the sensor (second bounce). Finally, the above model is used to analyze the field backscattered from a building as a function of the main scene parameters; in particular, the angle between vertical walls and sensor line of night and the dependence on the look angle are analyzed.

<bold>ABC</bold>s of foveal vision

This paper presents a simple mathematical performance model of the human foveal vision system based on an extensive analysis of the Blackwell-McCready (BM) data set. It includes a closed-form equation, the (ABC)t law, that allows the analyst to predict the entire range of BM threshold data. Relationships are derived among the four fundamental parameters of foveal vision: target area A, background luminance B, threshold contrast C, and stimulus presentation time t. Hyperbolic-curve fits on log-log plots of the data lead to the well-known laws of Ricco, Blackwell, Weber and Fechner, and Bloch. This paper unifies important relationships associated with target and background scene parameters as they relate to the human foveal vision process. The process of detecting a BM target, using foveal vision, is reduced to the total temporal summation of light energy modified by a multiplicative energy ratio. A stochastic model of human observer performance is presented in terms of a cumulative Gaussian distribution, which is a function of the apparent and BM contrast threshold values.

Age Differences in Perceptual Judgments about Potential Collision and Implications for Driving

To travel safely, drivers must detect and identify conditions that precipitate collisions. Previous results suggest that judgments about collision are influenced by scene parameters, task constraints, and age. Such studies typically focussed on judgments about when a collision would occur (time to contact). Results indicated that older observers underestimated time to contact more than younger observers; they thought that collisions would occur relatively sooner (Hancock & Manser, 1997; Schiff et al., 1992). However, biases toward underestimation would decrease the risk for accidents (Scialfa et al., 1987). Therefore, the age difference in judgments about time to contact does not seem to be a good candidate to account for the higher accident rate in older drivers. Alternatively, age differences in judgments about whether a collision would occur (potential collision; DeLucia, 1995) may contribute to differential accident rates. Therefore, age differences in such judgments were measured with computer simulations of three-dimensional scenes. When participants reported whether two moving objects would collide with each other, older adults (51-76 years) responded less accurately and more quickly than younger adults (18-23 years). When participants reported whether an approaching object would hit them, mean thresholds were greater for older females compared with younger females. The results indicate that the ability to detect collisions is an important factor to consider in an account of age differences in accident rates. Further research is warranted to determine whether less effective judgments about whether a collision would occur contribute more to the higher rate of traffic accidents and convictions in older drivers than to (mis)estimates of when a collision would occur.

Designing Virtual Environments to Enhance Human Performance

The current study focuses on design factors that can be manipulated to enhance human performance in virtual environment training systems (VETS). While there is currently an effort to understand for which tasks enhanced human performance can be achieved through VE training, there is limited understanding of how the VETS itself can be designed to enhance human performance in virtual environments. This study aims to establish a set of guiding design principles that enable intuitive and efficient interaction such that human performance in VETS is enhanced. This was accomplished by relating exposure duration, scene parameters, and amount of motion to human performance. A 3times2 between subjects experimental design was used, with 20 participants per treatment condition. The independent variables included exposure duration (15, 30, 45 minutes) and scene content (complex versus simple). The dependent variables included task performance, the amount of rotational roll, pitch and yaw, the amount of translational motion, as well as subjective reports of cybersickness taken immediately after VE exposure. The results were analyzed using non-parametric tests and correlational analyses. The results indicated that performance increased with amount of rotational motion (roll, pitch, and yaw) but was unrelated to the amount of translational motion. Scene content did not have a significant effect on performance nor sickness. In addition, sickness was not correlated with performance, nor, surprisingly, with amount of rotational or translational motion. Sickness did increase as exposure duration increased. The implication is that to maximize human performance, allow for sufficient roll, pitch, and yaw movement to accomplish task objectives, while ensuring the duration of exposure is less than 30 minutes.

ROBUST NEURAL APPROACH FOR THE ESTIMATION OF THE ESSENTIAL PARAMETERS IN COMPUTER VISION

The essential parameters approach is a well-known technique in computer vision for recovering the motion and scene parameters from a sequence of images. This approach has long been considered as suboptimal because of the underestimation of the effect of noise and outliers. This paper re-evaluates this method because it is correctly classified as a structured Total Least Squares problem and proposes very robust linear neurons for its solution. Then, a novel neural network, CASEDEL EXIN, which exploits these neurons together with the case deletion diagnostics, is introduced. It is not only very robust (outlier rejection), but is also able to identify the outliers. This fact can be exploited, for example, to refine the image segmentation.

Exploiting the generic viewpoint assumption

The “generic viewpoint” assumption states that an observer is not in a special position relative to the scene. It is commonly used to disqualify scene interpretations that assume special viewpoints, following a binary decision that the viewpoint was either generic or accidental. In this paper, we apply Bayesian statistics to quantify the probability of a view, and so derive a useful tool to estimate scene parameters. Generic variables can include viewpoint, object orientation, and lighting position. By considering the image as a (differentiable) function of these variables, we derive the probability that a set of scene parameters created a given image. This scene probability equation has three terms: the fidelity of the scene interpretation to the image data; the prior probability of the scene interpretation; and a new genericity term, which favors scenes likely to produce the observed image. The genericity term favors image interpretations for which the image is stable with respect to changes in the generic variables. It results from integration over the generic variables, using a low-noise approximation common in Bayesian statistics. This approach may increase the scope and accuracy of scene estimates. It applies to a range of vision problems. We show shape from shading examples, where we rank shapes or reflectance functions in cases where these are otherwise unknown. The rankings agree with the perceived values.

Method for estimating scene parameters from color histograms

One of the key tools in applying physics-based models to machine vision has been the analysis of color In the mid-1980’s it was recognized that the color histogram for a single inhomogeneous surface histograms. with highlights will have a planar distribution in color space. It has since been shown that the colors do not in a plane but form clusters at specific points. Physics-based models of reflection predict that fall randomly shape of the histogram is related not only to the illumination color and the object color but also to such the properties as surface roughness and imaging geometry. We present an algorithm for analyzing color noncolor histograms that yields estimates of surface roughness, phase angle between the camera and the light source, and illumination intensity. These three scene parameters are related to three histogram measurements. However, the relationship is complex and cannot be solved analytically. Therefore we developed a method for estimating these properties that is based on interpolation between histograms that come from images of known scene properties. We present tests of our algorithm on simulated data, and the results compare well with the known simulation parameters. We also test our method on real images, and the results compare favorably with the actual parameters estimated by other means. Our method for estimating scene properties is very fast and requires only a single color image.

An iterative inversion algorithm with application to the polarimetric radar response of vegetation canopies

The retrieval of scene parameters from polarimetric radar data using an iterative inversion approach is considered. The theoretical development of a general, model-based iterative algorithm for inversion of polarimetric radar data is presented. Factors relevant to its implementation, such as sensor configuration, algorithm optimization and computational structure are discussed. The algorithm is applied to the specific problem of inverting the vector radiative transfer model for a simplified, representative vegetation canopy consisting of vertical trunks, leaves, and a rough ground surface. The results of this inversion are in excellent agreement with simulated data generated using the radiative transfer model. The convergence properties of the algorithm are evaluated, and it is found that successful convergence is achieved in about 90% to 95% of the cases tested for the implementation used in this work. An error analysis is presented which considers the effect of both systematic and measurement derived errors. Typical error bounds for the current application are approximately /spl plusmn/3%, allowing for /spl plusmn/0.5 dB accuracy in the measured radar data.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Dynamic 3-D scene analysis through synthesis feedback control

The analysis of 3-D scenes consisting of nonrigid moving objects using 2-D image sequences is discussed. A parametric description of dynamic objects is extracted, and the time-variant scene parameters are estimated throughout the sequence by employing an analysis-by-synthesis approach. Images are synthesized from the parametric scene description and compared with the original images input to the camera. Frame differences between the synthesized and original images are evaluated to obtain an estimated scene parameter update. The analysis method is applied to videophone scenes in the form of a data compression algorithm, whereby the scene parameters are transmitted and the output sequence is synthesized at the receiver. >

The use of variograms in remote sensing: I. Scene models and simulated images

Extraction of information from remotely sensed images would greatly benefit from increased use of spatial data. However, the utility of spatial data has been undermined by a lack of understanding of the nature and causes of observed spatial variation in images. One approach to this problem is to model the spatial variation in images as a function of ground scene and sensor parameters. Variograms are the tool used to link models of ground scenes to spatial variation in images. Explicit variograms are calculated for simple models of ground scenes consisting of randomly located discs on a continuous background. By incorporating the effect of the IFOV of the sensor through a process called regularization, explicit variograms for images of these scene models are derived. Verification of the explicit variograms is accomplished by simulating images that match the assumed scene model and sensor parameters and calculating empirical variograms for these images. For a simple scene model of randomly located discs on a continuous background, explicit and empirical variograms match, verifying the convergence of these two dissimilar approaches. The sensitivity of variograms is studied through varying parameters of scene models both in calculating explicit variograms and in simulating images. Results indicate direct ties between several scene parameters and the behavior of variograms. The height of the sill of the variogram is related to the density of objects. The range of influence is related to the size of objects. Increased variance in the size distribution of objects results in a more rounded shape in the variogram near the sill. Primary effects of increasing the units of regularization, or larger pixel size, are a decrease in the height of the sill and an increase in the range of influence.

The influence of the scene parameters and of noise on the behaviour of automatic focusing algorithms

ABSTRACTA systematic study is presented of the properties of autofocus criteria. Special importance is attributed to their behaviour with respect to noise, working range and image sharpness, whose reproducibility has been investigated. It can be demonstrated that autofocus algorithms have either a large working range and do not focus in an unambiguous way or vice versa. From considerations of the properties of autofocus algorithms two novel algorithms have been developed which are insensitive to noise. One of these algorithms is characterized by a particularly large working range, whereas the other is capable of achieving maximum sharpness in a reproducible manner.

Texture, contour, shape, and motion

Intrinsic magic calculation exploits constraints arising from physical and imaging processes to derive physical scene parameters from input images. After a brief review of a paradigmatic intrinsic image calculation we turn to a recent result in shape from texture and then to a new result that derives shape and motion from a sequence of patterned inputs. Experimental results are demonstrated for synthetic and natural images.

Use of multiple difference-of-Gaussian filters to verify geometric models

Experiments are reported with a program which recognizes London buses in busy street scenes. The program uses a 3D model of a bus. together with knowledge of camera and scene parameters, to compute the visual appearance of possible target events. In earlier work this took the form of 2D templates, which identify locations within the image where abrupt discontinuities of grey-value are expected. Each template corresponds to an hypothesis which is verified by measuring the agreement between the predicted features and the response of a Sobel filter to the image data. This paper reports an extension to the program which offers improved selectivity, and a more general-purpose method of linking predicted features to detected imageattributes. Multiple difference-of-Gaussian filters perform an initial object-independent analysis of the image to provide an intermediate domain in which object-dependent features, predicted by hypothesized instances of the model, are tested.

TV requirements for manipulation in space

An analytical and experimental study was conducted to specify a video system for remote manipulation in space. An operator function analysis identified two basic characteristics, work volume and element relationship, which define 4 manipulation tasks chosen for examination. A visual function analysis developed a set of elemental scene parameters which grouped the visual dimensions into major areas of influence. Simulation testing was conducted with a 4 degree-of-freedom motion frame which allowed an operator to perform the manipulation tasks. Four video systems were included in the simulation testing: a black and white and a color monoscopic system, a stereoscopic system, and a black and white 2-view system. A sequential experimental plan first provided an overall analysis of the effects of tasks, scene parameters and video systems. This was followed by a detailed experimental examination of the critical dimensions identified in the first experiment. Results are discussed in terms of a recommended TV system.

Television Systems for Remote Manipulation

An analytical and experimental study was conducted to specify a video system for remote manipulation in space. An operator function analysis identified two basic characteristics, work volume and element relationship, which define four manipulation tasks chosen for examination. A visual function analysis developed a set of elemental scene parameters which grouped the visual dimensions into major areas of influence. Simulation testing was conducted with a four degree-of-freedom motion frame which allowed an operator to perform the manipulation tasks. Four video systems were included in the simulation testing: a black and white and a color mono-scopic system, a stereoscopic system, and a black and white two-view system. A sequential experimental plan first provided an overall analysis of the effects of tasks, scene parameters, and video systems. This was followed by a detailed experimental examination of the critical dimensions identified in the first experiment. Results are discussed in terms of a recommended TV system.