Engineering topics

Ahumada, A. J., Jr.

Publications and source records attributed to Ahumada, A. J., Jr..

13 records

NASA NTRS2000

Image Discrimination Models for Object Detection in Natural Backgrounds

This paper reviews work accomplished and in progress at NASA Ames relating to visual target detection. The focus is on image discrimination models, starting with Watson's pioneering development of a simple spatial model and progressing through this model's descendents and extensions. The application of image discrimination models to target detection will be described and results reviewed for Rohaly's vehicle target data and the Search 2 data. The paper concludes with a description of work we have done to model the process by which observers learn target templates and methods for elucidating those templates.

Ahumada, A. J., Jr.↗

NASA NTRS1998

What Determines the Response: Test or Reference?

The stability of sensory memory has been studied by presenting a reference stimulus, a delay, and a test stimulus. As has been pointed out by Lages and Treisman (1998 Vision Research 38 557-572), the usual measure of performance depends only on the effect of test variations on the responses. The Weber fraction characterizing performance is more properly called the test stimulus Weber fraction. We measure the relative contribution of the test and reference to the response by the ratio of the test Weber fraction to the reference Weber fraction. The stimuli were two dark lines on a bright background. Seven reference separations, varying from 9.5 to 16.7 arc min, were intermixed in each run. Interstimulus intervals (ISI) of 50, 200 and 2000 msec and intertrial intervals (ITI) of 500 and 2500 msec were investigated. When the ISI was short (50 or 200 msec), for both ITIs, responses were determined equally by the test and reference. For the long ISI (2000 msec), the reference stimulus contributed less. However, only for the 500 msec ITI (and not for all observers) was the contribution of the reference stimulus negligible, as Treisman's criterion setting theory might suggest.

Chukova, S. V.↗

NASA NTRS1995

Object Detection in Natural Backgrounds Predicted by Discrimination Performance and Models

In object detection, an observer looks for an object class member in a set of backgrounds. In discrimination, an observer tries to distinguish two images. Discrimination models predict the probability that an observer detects a difference between two images. We compare object detection and image discrimination with the same stimuli by: (1) making stimulus pairs of the same background with and without the target object and (2) either giving many consecutive trials with the same background (discrimination) or intermixing the stimuli (object detection). Six images of a vehicle in a natural setting were altered to remove the vehicle and mixed with the original image in various proportions. Detection observers rated the images for vehicle presence. Discrimination observers rated the images for any difference from the background image. Estimated detectabilities of the vehicles were found by maximizing the likelihood of a Thurstone category scaling model. The pattern of estimated detectabilities is similar for discrimination and object detection, and is accurately predicted by a Cortex Transform discrimination model. Predictions of a Contrast- Sensitivity- Function filter model and a Root-Mean-Square difference metric based on the digital image values are less accurate. The discrimination detectabilities averaged about twice those of object detection.

Ahumada, A. J., Jr.↗

NASA NTRS1994

The Visibility of DCT Quantization Noise: Spatial Frequency Summation

Computational models of the ability to detect image compression artifacts facilitate the optimization of compression parameters, such as the JPEG quantization matrix. For simplicity, some models assume that the visibility of artifacts containing different spatial frequency components is determined by the most visible component, that is, no summation over components. Using the type of noise generated by quantization in the Discrete Cosine Transform (DCT) domain, we find a degree of summation between that of probability summation and contrast energy summation.

Peterson, H. A.↗

NASA NTRS1994

Smoothing DCT Compression Artifacts

Image compression based on quantizing the image in the discrete cosine transform (DCT) domain can generate blocky artifacts in the output image. It is possible to reduce these artifacts and RMS error by adjusting measures of block edginess and image roughness, while restricting the DCT coefficient values to values that would have been quantized to those of the compressed image. We also introduce a DCT coefficient amplitude adjustment that reduces RMS error.

Ahumada, A. J., Jr.↗

NASA NTRS1986

Window of visibility - A psychophysical theory of fidelity in time-sampled visual motion displays

A film of an object in motion presents on the screen a sequence of static views, while the human observer sees the object moving smoothly across the screen. Questions related to the perceptual identity of continuous and stroboscopic displays are examined. Time-sampled moving images are considered along with the contrast distribution of continuous motion, the contrast distribution of stroboscopic motion, the frequency spectrum of continuous motion, the frequency spectrum of stroboscopic motion, the approximation of the limits of human visual sensitivity to spatial and temporal frequencies by a window of visibility, the critical sampling frequency, the contrast distribution of staircase motion and the frequency spectrum of this motion, and the spatial dependence of the critical sampling frequency. Attention is given to apparent motion, models of motion, image recording, and computer-generated imagery.

Watson, A. B.↗

NASA NTRS1985

Application of a computable model of human spatial vision to phase discrimination

A computable model of human spatial vision is used to make predictions for phase-discrimination experiments. This model is being developed to deal with a broad range of problems in vision and was not specifically formulated to deal with phase discrimination. In the model, cross-correlation of the stimuli with an array of sensors produces feature vectors that are operated on by a position-uncertain ideal observer to simulate detection and discrimination experiments. In this report, the stimuli are compound sinusoidal gratings composed of a fundamental and a higher-frequency component added in various phases. Model predictions are compared with three key results from the literature: (1) the effect of the contrast of the fundamental on phase discrimination, (2) threshold phase difference as a function of the fundamental frequency, and (3) the contrast required for phase discrimination as a function of the frequency ratio of the two grating components. In the first two cases, the predictions capture the main features of the data, although quantitative discrepancies remain. In the third case, the model fails, and this failure suggests additional restrictions on the combination of information across sensors.

Nielsen, K. R. K.↗

NASA NTRS1985

Equivalent-noise model for contrast detection and discrimination

Models for the detection and the discrimination of low-contrast signals by human observers typically assume that the observer is limited by the filtering action of the visual system and by the noisy character of its processing. For some models both the filtering and the noise can be represented by a noise in the stimulus domain, the input equivalent noise of the model. The present study derives some formulas for computing this noise, describes the calculation of a sample, and discusses some implications of this approach.

Ahumada, A. J., Jr.↗

NASA NTRS1985

Model of human visual-motion sensing

A model of how humans sense the velocity of moving images is proposed. The model exploits constraints provided by human psychophysics, notably that motion-sensing elements appear tuned for two-dimensional spatial frequency, and by the frequency spectrum of a moving image, namely, that its support lies in the plane in which the temporal frequency equals the dot product of the spatial frequency and the image velocity. The first stage of the model is a set of spatial-frequency-tuned, direction-selective linear sensors. The temporal frequency of the response of each sensor is shown to encode the component of the image velocity in the sensor direction. At the second stage, these components are resolved in order to measure the velocity of image motion at each of a number of spatial locations and spatial frequencies. The model has been applied to several illustrative examples, including apparent motion, coherent gratings, and natural image sequences. The model agrees qualitatively with human perception.

Watson, A. B.↗

NASA NTRS1984

Uniform apparent contrast noise: A picture of the noise of the visual contrast detection system

A picture which is a sample of random contrast noise is generated. The noise amplitude spectrum in each region of the picture is inversely proportional to spatial frequency contrast sensitivity for that region, assuming the observer fixates the center of the picture and is the appropriate distance from it. In this case, the picture appears to have approximately the same contrast everywhere. To the extent that contrast detection thresholds are determined by visual system noise, this picture can be regarded as a picture of the noise of that system. There is evidence that, at different eccentricities, contrast sensitivity functions differ only by a magnification factor. The picture was generated by filtering a sample of white noise with a filter whose frequency response is inversely proportional to foveal contrast sensitivity. It was then stretched by a space-varying magnification function. The picture summmarizes a noise linear model of detection and discrimination of contrast signals by referring the model noise to the input picture domain.

Ahumada, A. J., Jr.↗

NASA NTRS1983

The window of visibility: A psychological theory of fidelity in time-sampled visual motion displays

Many visual displays, such as movies and television, rely upon sampling in the time domain. The spatiotemporal frequency spectra for some simple moving images are derived and illustrations of how these spectra are altered by sampling in the time domain are provided. A simple model of the human perceiver which predicts the critical sample rate required to render sampled and continuous moving images indistinguishable is constructed. The rate is shown to depend upon the spatial and temporal acuity of the observer, and upon the velocity and spatial frequency content of the image. Several predictions of this model are tested and confirmed. The model is offered as an explanation of many of the phenomena known as apparent motion. Finally, the implications of the model for computer-generated imagery are discussed.

Watson, A. B.↗

NASA NTRS1983

A look at motion in the frequency domain

A moving image can be specified by a contrast distribution, c(x,y,t), over the dimensions of space x,y, and time t. Alternatively, it can be specified by the distribution C(u,v,w) over spatial frequency u,v and temporal frequency w. The frequency representation of a moving image is shown to have a characteristic form. This permits two useful observations. The first is that the apparent smoothness of time-sampled moving images (apparent motion) can be explained by the filtering action of the human visual system. This leads to the following formula for the required update rate for time-sampled displays. W(c)=W(l)+ru(l) where w(c) is the required update rate in Hz, W(l) is the limit of human temporal resolution in Hz, r is the velocity of the moving image in degrees/sec, and u(l) is the limit of human spatial resolution in cycles/deg. The second observation is that it is possible to construct a linear sensor that responds to images moving in a particular direction. The sensor is derived and its properties are discussed.

Watson, A. B.↗

NASA NTRS1983

Reduction of display artifacts by random sampling

The application of random-sampling techniques to remove visible artifacts (such as flicker, moire patterns, and paradoxical motion) introduced in TV-type displays by discrete sequential scanning is discussed and demonstrated. Sequential-scanning artifacts are described; the window of visibility defined in spatiotemporal frequency space by Watson and Ahumada (1982 and 1983) and Watson et al. (1983) is explained; the basic principles of random sampling are reviewed and illustrated by the case of the human retina; and it is proposed that the sampling artifacts can be replaced by random noise, which can then be shifted to frequency-space regions outside the window of visibility. Vertical sequential, single-random-sequence, and continuously renewed random-sequence plotting displays generating 128 points at update rates up to 130 Hz are applied to images of stationary and moving lines, and best results are obtained with the single random sequence for the stationary lines and with the renewed random sequence for the moving lines.

Ahumada, A. J., Jr.↗