The measurement of the salience of targets and distractors through competitive novelty
F W M Stentiford (UCL@Adastral Park, University College London, Adastral Park,
Martlesham, Ipswich IP5 3RE, UK; e-mail:f.stentiford@ee.ucl.ac.uk)
Salience is normally strongest in a region if small neighbourhoods in that region are
dissimilar to neighbourhoods elsewhere in the image. Measures based on this principle
have been used for image compression [1] and assessing cell damage [2]. This paper
reports results of experiments using images of targets and distractors [3,4,5] in which the
roles of colour, orientation and scale are explored. It is found that the edges of objects
and line ends attract high measures of attention unless those edges and line ends
themselves predominate in the image. Distractors that obtain higher measures of attention
normally correspond to those cases in which targets are harder to find. For example, the
higher measures of attention are generated by distractors when they are different. When
orientation differences are small the neighbourhoods must be of a sufficient size to
encompass features that discriminate between target and distractors. This draws a parallel
with the finding that smaller-group search is encouraged by densely packed arrays [3].
[1]
[2]
[3]
[4]
[5]
F W M Stentiford, “An estimator for visual attention through competitive novelty
with application to image compression, 22ndPicture Coding Symposium, April
2001.
F W M Stentiford, “Automatic identification of regions of interest with
application to the quantification of DNA damage in cells,”, Proc. SPIE Vol. 4662,
pp 244-253, 2002.
K V Sobel & K R Cave, “Roles of salience and strategy in conjunction search”, J.
Experimental Psychology, Human Perception and Performance, Vol 28, No 5, pp
1055-1070, 2002.
J M Wolfe et al, “Segmentation of objects from backgrounds in visual search
tasks,” Vision Research, 42, pp 2985-3004, 2002.
J M Wolfe, “Visual Search”, in Attention, H Pashler editor, Psyschology Press,
1998.