File Download
There are no files associated with this item.
Supplementary
-
Citations:
- Appears in Collections:
Conference Paper: The detection of smooth curves in jagged contours
Title | The detection of smooth curves in jagged contours |
---|---|
Authors | |
Issue Date | 2002 |
Publisher | Pion Ltd.. The Journal's web site is located at http://www.perceptionweb.com |
Citation | The 25th European Conference on Visual Perception, Glasgow, Scotland, 25-29
August 2002. In Perception, ECVP Abstract, 2002, v. 31 n. Suppl., p. 153-154 How to Cite? |
Abstract | Jagged contours with sharp edges, such as those formed by the fractal edge of the Mandelbrot set, have
a lower correlation in position and shape between different scales than smooth curves. The more jagged
the contour, the lower the correlation. In principle, coarse changes in position of a curve, be it jagged
or smooth, can be coded by using operators tuned to low-luminance spatial frequencies; but, to code
the fine changes in position of a jagged contour, small operators are required. Since a sharp-luminance-
transition contour is visible to both high-spatial-frequency-tuned and low-spatial-frequency-tuned
operators, the question arises whether coarse changes in shape are coded via low-spatial-frequency-
tuned cells, or via high-spatial-frequency-tuned cells whose outputs are pooled by a coarse-scale
higher-order mechanism. We investigated this issue by conducting two experiments. In the first
experiment, we show that shape-amplitude detection thresholds of a sinusoidal contour increase with
the blur of the contour, and that the threshold functions are parallel over a large range of shape
spatial frequencies, implying that high-(luminance)-spatial-frequency-tuned mechanisms that are
more shape-and-position sensitive come into play whenever they can be supported. In the second experiment, we measured the threshold for shape-amplitude detection of a low-shape-frequency
sinusoidal contour in the presence of shape harmonics of various amplitudes, for both (luminance)
spatial-frequency unfiltered, and high-pass spatial-frequency filtered, contours.We found no differ-
ence in thresholds for shape-amplitude detection when using filtered and unfiltered stimuli. We
conclude that coarse shape information is coded with the smallest operators available. In the
context of a model of vision that emphasises the role of self-similar `filters' tuned in spatial
frequency, our results are consistent with the idea that contour shape is encoded via high-spatial-
frequency-tuned cells, possibly of the highest spatial frequency capable of responding to the contour.
[Supported by URC (Hong Kong) and NSERC (Canada).] |
Description | Oral presentations: Perceptual organisation |
Persistent Identifier | http://hdl.handle.net/10722/197959 |
ISSN | 2023 Impact Factor: 1.6 2023 SCImago Journal Rankings: 0.584 |
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Hayes, A | - |
dc.contributor.author | Kingdom, FAA | - |
dc.contributor.author | Prins, N | - |
dc.date.accessioned | 2014-06-17T02:31:40Z | - |
dc.date.available | 2014-06-17T02:31:40Z | - |
dc.date.issued | 2002 | - |
dc.identifier.citation | The 25th European Conference on Visual Perception, Glasgow, Scotland, 25-29 August 2002. In Perception, ECVP Abstract, 2002, v. 31 n. Suppl., p. 153-154 | - |
dc.identifier.issn | 0301-0066 | - |
dc.identifier.uri | http://hdl.handle.net/10722/197959 | - |
dc.description | Oral presentations: Perceptual organisation | - |
dc.description.abstract | Jagged contours with sharp edges, such as those formed by the fractal edge of the Mandelbrot set, have a lower correlation in position and shape between different scales than smooth curves. The more jagged the contour, the lower the correlation. In principle, coarse changes in position of a curve, be it jagged or smooth, can be coded by using operators tuned to low-luminance spatial frequencies; but, to code the fine changes in position of a jagged contour, small operators are required. Since a sharp-luminance- transition contour is visible to both high-spatial-frequency-tuned and low-spatial-frequency-tuned operators, the question arises whether coarse changes in shape are coded via low-spatial-frequency- tuned cells, or via high-spatial-frequency-tuned cells whose outputs are pooled by a coarse-scale higher-order mechanism. We investigated this issue by conducting two experiments. In the first experiment, we show that shape-amplitude detection thresholds of a sinusoidal contour increase with the blur of the contour, and that the threshold functions are parallel over a large range of shape spatial frequencies, implying that high-(luminance)-spatial-frequency-tuned mechanisms that are more shape-and-position sensitive come into play whenever they can be supported. In the second experiment, we measured the threshold for shape-amplitude detection of a low-shape-frequency sinusoidal contour in the presence of shape harmonics of various amplitudes, for both (luminance) spatial-frequency unfiltered, and high-pass spatial-frequency filtered, contours.We found no differ- ence in thresholds for shape-amplitude detection when using filtered and unfiltered stimuli. We conclude that coarse shape information is coded with the smallest operators available. In the context of a model of vision that emphasises the role of self-similar `filters' tuned in spatial frequency, our results are consistent with the idea that contour shape is encoded via high-spatial- frequency-tuned cells, possibly of the highest spatial frequency capable of responding to the contour. [Supported by URC (Hong Kong) and NSERC (Canada).] | - |
dc.language | eng | - |
dc.publisher | Pion Ltd.. The Journal's web site is located at http://www.perceptionweb.com | - |
dc.relation.ispartof | Perception | - |
dc.title | The detection of smooth curves in jagged contours | en_US |
dc.type | Conference_Paper | en_US |
dc.identifier.email | Hayes, A: ahayes@hkucc.hku.hk | - |
dc.identifier.doi | 10.1068/v020430 | - |
dc.identifier.hkuros | 108628 | - |
dc.identifier.hkuros | 82930 | - |
dc.identifier.volume | 31 | - |
dc.identifier.issue | Suppl. | - |
dc.identifier.spage | 153 | - |
dc.identifier.epage | 154 | - |
dc.publisher.place | United Kingdom | - |
dc.identifier.issnl | 0301-0066 | - |