saccade: A movement of the eye or camera, changing the direction of fixation sharply. [WP:Saccade]
salience: The extent to which something (e.g., a visual feature) stands out relative to other nearby features (from the Latin salire meaning to leap). [Itt01]
saliency map: A representation encoding the saliency of given image elements, typically features or groups thereof. See also salient feature, Gestalt, perceptual grouping and perceptual organization. [WP:Salience_(neuroscience)]
salient behavior: Behavior of a person or system which is distinct from normal behavior.
salient feature: A feature associated with a high value of a saliency measure, quantifying feature suggestiveness for perception. For instance, inflection points have been indicated as salient features for representing contours. Saliency is a concept that originated from Gestalt psychology. See also perceptual grouping and perceptual organization. [KK98]
salient pixel group: A group of pixels that exhibits a distinct pattern relative to neighboring pixels. [XG06]
salient point: Typically, a feature which is distinct relative to those around it. [SL03]
salient regions: Image regions that are interesting relative to their local image context. They should be stable to global transformations (including scale, illumination and perspective distortions) and image noise. They can be used for object representation, correspondence matching, tracking etc. [KZB04]
salt-and-pepper noise: A type of impulsive noise. Let
x,
y [0, 1] be two uniform random variables,
I the true image value at a given pixel and
In the corrupted (noisy) version of
I. We can define the effect of salt-and-pepper noise as
In =
imin +
y (
imax −
imin) iff
x ≥
l, where
l is a parameter controlling how much of the image is corrupted and
imin,
imax are the range of the noise. See also
image noise and
Gaussian noise. The figure was corrupted with 1% noise: [TV98:3.1.2]
sample covariance: For a
d-dimensional data set represented as a set of
n column vectors
for
i= 1, …,
n with
sample mean , the sample covariance is the
d ×
d matrix
. See also
covariance matrix. [MKB79:1.4.1]
sample mean: For a
d-dimensional data set represented as a set of
n column vectors
for
i= 1, …,
n, the sample mean is
. See also
mean. [MKB79:1.4.1]
sampling: The transformation of a continuous signal into a discrete one by recording its values at discrete instants or locations. Most digital images are sampled in space, time and intensity, as intensity values are defined only on a regular spatial grid and can only take integer values. The figure shows a continuous signal and its samples: [FP03:7.4.1]
sampling bias: If samples are collected from a random variable according to the true distribution then any statistic computed from the sample should not deviate systematically from the population expectation. If the sample does not represent the true distribution there is said to be “sampling bias”. [WP:Bias_(statistics)]
sampling density: The density of a sampling grid, that is, the number of samples collected per unit interval. See also sampling. [BB82:2.2.6]
sampling theorem: If an image is sampled at a rate higher than its Nyquist fre...
