PREATTENTIVE TEXTURE SEGMENTATION - THE ROLE OF LINE TERMINATIONS, SIZE, AND FILTER WAVELENGTH

Citation
Bs. Rubenstein et D. Sagi, PREATTENTIVE TEXTURE SEGMENTATION - THE ROLE OF LINE TERMINATIONS, SIZE, AND FILTER WAVELENGTH, Perception & psychophysics, 58(4), 1996, pp. 489-509
Citations number
35
Categorie Soggetti
Psychology, Experimental",Psychology
Journal title
ISSN journal
00315117
Volume
58
Issue
4
Year of publication
1996
Pages
489 - 509
Database
ISI
SICI code
0031-5117(1996)58:4<489:PTS-TR>2.0.ZU;2-0
Abstract
The triangle/(sic) texel pair represents a dilemma in texture discrimi nation because, despite having the same component orientations, discri mination is still possible (Julesz, 1981), showing a performance asymm etry. Other possible element properties that could influence this task are line terminations, closure, and the size of these elements. We fo und that line terminators are critical for the task; however, results from double-task experiments indicated that terminator-based discrimin ation requires the use of attention. When attention is not available f or the task, ''size'' of the elements (with the (sic) considered sligh tly larger) seems to be critical for this discrimination and for the a symmetric performance. To generalize the concept of ''size'' to textur es in general, further experiments were performed with textures of dif ferent-sized elements. Results showed, as past literature has indicate d, that there is a performance asymmetry, with the larger of the eleme nts being more visible when in the foreground. This asymmetry was addi tionally shown to reverse itself (i.e., the smaller element became the more visible) as the scale of the elements increased (while interelem ent distance remained fixed). A filter analysis was developed in order to measure the apparent size of these elements within textures (texsi ze), defined as the response weighted average of the filter wavelength , <(lambda)over bar>, for a group of elements. The calculation of <(la mbda)over bar> was attained by introducing a nonlinearity after the se cond stage of filtering (or spatial averaging of filter responses). Th is analysis showed high correlation between the texture with the large r <(lambda)over bar> and the more visible texture. On the basis of thi s correlation, a wavelength-dependent noise is proposed, having more i nternal noise for low-spatial-frequency filters and less for high-spat ial-frequency filters.