Expand this Topic clickable element to expand a topic
Skip to content
Optica Publishing Group

Reconstruction of temporally undersampled images using motion direction information

Open Access Open Access

Abstract

Several problems posed by spatial and temporally sampled moving images may be ameliorated when knowledge of visual motion direction is known. We extend prior analyses of sampled moving images through the following conjecture. For rigid motion in one direction, the image spatiotemporal Fourier transform is a surface which occupies only two diagonal quadrants of the spatiotemporal frequency plane. Although sampling introduces energy to all four quadrants, that which falls over the wrong diagonal quadrants is likely to be more detrimental to motion perception, since the energy tends to excite motion detectors signaling an incorrect direction. If the direction of motion is known at every point in an image, temporally undersampled images may be partially reconstructed by removing this incorrect energy. This conjecture serves to explain the demise of short-range motion perception at large displacements and interstimulus intervals. It also explains motion capture phenomena, cooperativity, and the coherence of sampled motion displays containing elements which are effectively undersampled.

© 1988 Optical Society of America

PDF Article
More Like This
Two-dimensional motion analysis: Information In the nodes of moving plaids

John A. Perrone and Leland S. Stone
THS5 OSA Annual Meeting (FIO) 1988

Deep learning-based image reconstruction in optical coherence tomography using undersampled spectral data

Yijie Zhang, Tairan Liu, Manmohan Singh, Ege Çetintaş, Yilin Luo, Yair Rivenson, Kirill V. Larin, and Aydogan Ozcan
AM5I.7 CLEO: Applications and Technology (CLEO:A&T) 2022

New phenomena show effects of luminance motion on stereoscopic motion and on color motion

Jih Jie Chang and Bela Julesz
THEE1 OSA Annual Meeting (FIO) 1988

Select as filters


Select Topics Cancel
© Copyright 2024 | Optica Publishing Group. All rights reserved, including rights for text and data mining and training of artificial technologies or similar technologies.