Open Collections

Selection and organization of subjective contours

University of British Columbia

Subjective contours are physically invisible borders drawn on certain images that can nevertheless be seen by humans. This is because the human vision system makes assumptions on the occlusion of objects. The study of subjective contours is important for helping us understand more about the human visual perception. The purpose of this thesis is to understand the perception of subjective contours and to detect subjective contours by computer. The previous subjective contour detection systems limit the subjective contours they can detect by restricting the locations on the figures where the subjective contours can be seen and by using the consistent subjective surface orientation. In this thesis, we consider the overall organization of subjective contours. We do not put the restriction on the subjective surface orientation because we view the subjective contour as a boundary separating the two regions locally. A model for subjective contour detection is presented based on four criteria: no prior knowledge is necessary to detect a subjective contour; a subjective contour is a special type of occluding contour; the shape of a subjective contour is determined by the viewing condition; and it is possible to have many subjective contour organizations from one image. The rules for subjective contour organization are described and the model explains different types of subjective contour organizations. There are three stages in the computer implementation of subjective contour detection. The first stage is preprocessing of figures where the real contours are segmented according to their curvature discontinuities by Lowe's curve partition method. The next stage is local processing in which each real contour segment selects all the potential subjective contours and their connecting real contour segments. The final stage is global processing to organize the real and subjective contours which can be seen at the same time. Many subjective contour images are tested and good results are produced.

Thesis/Dissertation

application/pdf

2009-02-26

For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.

http://hdl.handle.net/2429/5110

Computer Science

University of British Columbia

UBCV

DSpace