Global and local image features
7 important questions on Global and local image features
What properties should feature vectors of a global image descriptor be in order to allow comparing under different sized images?
Name an approach to allow comparing images with a variable number of local image features.
What local image feature descriptors does the "Bag of Words" aproach use, and how is this converted into a comparable (fixed) feature vector?
- Higher grades + faster learning
- Never study anything twice
- 100% sure, 100% understanding
What is a local image descriptor?
What are three important things in consideration of the number and quality of codewords for available local image features?
- The number of codewords should be sufficiently many in such that classes can be distinguished by them.
- The number of codewords should be limited general in such that resulting codeword-histograms are dense.
- Codewords should be sufficiently discriminative to be able to distinguish classes.
What role can K-Means play in the process of converting local image descriptors into classifying images?
What are the five steps for the Bag of Words approach?
- Find keypoints in image
- Calculate local feature descriptors (with e.g. SIFT)
- Map local descriptors to codewords by computing distance to codebook cluster centers.
- Normalize the BoW histogram to unit length.
The question on the page originate from the summary of the following study material:
- A unique study and practice tool
- Never study anything twice again
- Get the grades you hope for
- 100% sure, 100% understanding
![Trustpilot Logo](/images/design_2020/lp/clp-trustpilot-en.png)