PubMed Health⌕ Search
Home/Archive/PubMed

PubMed · 8577574

Size constancy in structure from motion.

J Turner·M L Braunstein
View original source ↗Export metadata ↓

Abstract

The relative motions of points in a structure-from-motion display involving parallel projection provide depth information in an object-centered framework: differences in velocity do not reflect differences in distance from an eyepoint. In contrast, size constancy is generally regarded to be a perspective effect, based on the relationship between projected size and distance from an eyepoint. Five subjects judged the relative sizes of objects in structure-from-motion scenes. Although the scenes were displayed without perspective, judged size was related to the simulated separation in depth of the objects. These results suggest that relative depths recovered from object-centered information are incorporated into a viewer-centered framework.

Explore related subjects

Depth PerceptionHumansMotion PerceptionJournal ArticleResearch Support, U.S. Gov't, Non-P.H.S.

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

J Turner, M L Braunstein. 1995. Size constancy in structure from motion.. https://doi.org/10.1068/p241155

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related citations

PubMed1999

Variational learning in nonlinear gaussian belief networks.

We view perceptual tasks such as vision and speech recognition as inference problems where the goal is to estimate the posterior distribution over latent variables (e.g., depth in stereo vision) given the sensory input. The recent flurry of research in independent component analysis exemplifies the importance of inferring the continuous-valued latent variables of input data. The latent variables found by this method are linearly related to the input, but perception requires nonlinear inferences such as classification and depth estimation. In this article, we present a unifying framework for stochastic neural networks with nonlinear latent variables. Nonlinear units are obtained by passing the outputs of linear gaussian units through various nonlinearities. We present a general variational method that maximizes a lower bound on the likelihood of a training set and give results on two visual feature extraction problems. We also show how the variational method can be used for pattern classification and compare the performance of these nonlinear networks with other methods on the problem of handwritten digit recognition.

Depth Perception↗
PubMed1998

Two-dimensional matches from one-dimensional stimulus components in human stereopsis.

Three-dimensional visual scenes project onto the retina of the eye as two-dimensional images. The third dimension, depth, is projected as subtle differences between left and right retinal images. As early as the 1830s, stereoscopic depth perception was shown to depend on horizontal disparities between these images. To detect disparity, the visual system must match corresponding parts of the two retinal images. To identify the stimulus elements used in stereo matching, I applied a disparity-adaptation technique to visual patterns whose one-dimensional components and two-dimensional features have very different disparities. Surprisingly, the adaptors that are effective in altering depth perception appear widely separated in depth from the patterns they adapt. I conclude that stereo matching occurs in all directions of two-dimensional space and that one-dimensional components are the stimulus primitives, the fundamental elements of stereo matching. This is a reversal of the classical view of stereo correspondence as a one-dimensional (horizontal) matching of monocular two-dimensional features.

Depth Perception↗