Hello, I am a bit confusing about the meaning of the terms, Image descriptor, feature descriptor, feature vector.
My understanding is illstrated as follow:
Image descriptor --> data get from the whole image.
Feature descriptor --> data get from the local point of an image(a small region).
Feature vector --> a bunch of data which could feed to the machine learning algorithm.
sometimes feature descriptor = feature vector because they could feed to the machine learning algorithm and extract from the local point of an image.
please correct me if I get them wrong.
This is a good question.
Here are the definitions:
image descriptor: An image descriptor is a representation of homogeneous features in an image or sub-image. For more about this, see
A. Ardeshir Goshtasby
Image Descriptors
Chapter Image Descriptors
feature descriptor: The term feature descriptor is sometimes used interchangeably with the term image descriptor. However, a feature descriptor is not restricted to homogenous image features.
In computer vision, a feature descriptor encodes such things as image region shape or holes (dark, light-absorbing image regions) or pixel colour or, importantly, pixel gradient or image texture or a Rosenfeld 4- or 8-neighbourhood.
For more about this, see
Book Foundations of Computer Vision
and
Book Topology of Digital Images. Visual Pattern Discovery in Prox...
feature vector: For an understanding of the term feature vector, we need to consider what is known as a feature space. Typically, a feature space is the collection of points in n-dimensional Euclidean space (denoted by Rn). A point in Euclidean space is specified with a tuple of real numbers such (x,y,z) for a point in R3. Each point in a feature space is specified by a tuple of real numbers, where each number in the tuple is a feature value. For example, a feature vector in R5 has the following form with respect to features
contour length (e.g., 220 mm),
shape interior area (e.g., 15 m2),
average green pixel intensity (e.g., 529 nm),
average line segment length (e.g., 2 mm),
edge pixel gradient angle (e.g., 55 deg.)
with the corresponding feature vector:
(220, 15, 529, 2, 55).
For more about Euclidean space, see
http://mathworld.wolfram.com/EuclideanSpace.html
For a detailed view of feature spaces, see
http://pages.cs.wisc.edu/~bsettles/cs540/lectures/16_feature_spaces.pdf
For more about feature vectors, see
https://www.pyimagesearch.com/2014/03/03/charizard-explains-describe-quantify-image-using-feature-vectors/
This is a good question.
Here are the definitions:
image descriptor: An image descriptor is a representation of homogeneous features in an image or sub-image. For more about this, see
A. Ardeshir Goshtasby
Image Descriptors
Chapter Image Descriptors
feature descriptor: The term feature descriptor is sometimes used interchangeably with the term image descriptor. However, a feature descriptor is not restricted to homogenous image features.
In computer vision, a feature descriptor encodes such things as image region shape or holes (dark, light-absorbing image regions) or pixel colour or, importantly, pixel gradient or image texture or a Rosenfeld 4- or 8-neighbourhood.
For more about this, see
Book Foundations of Computer Vision
and
Book Topology of Digital Images. Visual Pattern Discovery in Prox...
feature vector: For an understanding of the term feature vector, we need to consider what is known as a feature space. Typically, a feature space is the collection of points in n-dimensional Euclidean space (denoted by Rn). A point in Euclidean space is specified with a tuple of real numbers such (x,y,z) for a point in R3. Each point in a feature space is specified by a tuple of real numbers, where each number in the tuple is a feature value. For example, a feature vector in R5 has the following form with respect to features
contour length (e.g., 220 mm),
shape interior area (e.g., 15 m2),
average green pixel intensity (e.g., 529 nm),
average line segment length (e.g., 2 mm),
edge pixel gradient angle (e.g., 55 deg.)
with the corresponding feature vector:
(220, 15, 529, 2, 55).
For more about Euclidean space, see
http://mathworld.wolfram.com/EuclideanSpace.html
For a detailed view of feature spaces, see
http://pages.cs.wisc.edu/~bsettles/cs540/lectures/16_feature_spaces.pdf
For more about feature vectors, see
https://www.pyimagesearch.com/2014/03/03/charizard-explains-describe-quantify-image-using-feature-vectors/
- Badges
- Science topic
- Similar topics
- Computer Science
- Images