Advantages/Disadvantages

8 important questions on Advantages/Disadvantages

BoF Advantages/Disadvantages

+     Translation invariant
+     Robust matching of visual word instances
+     Fixed length feature vector
+     Good performance (simplicity)

-     Loss of spatial position and object configuration (so, not suitable for recognise or localise images)
-     Quantisation loses information

Cross-correlation Advantages/Disadvantages

+     Linear & shift invariant
+     Commutative (F * G = G * F)

-     Difficult when working with the edge:
     +     Ignore
     +     Mirroring
     +     Wrap around
     +     Copy edge (assume image continues like on the edge)

Gradient/Image derivative Advantages/Disadvantages

+     Robust feature and texture matching

-     Different lightning or camera position can cause 2 images of the same scene to have drastically different pixel values
     +     Compute texture or feature signatures based on gradient images computed from the original images.
  • Higher grades + faster learning
  • Never study anything twice
  • 100% sure, 100% understanding
Discover Study Smart

Harris Corner Detection Advantages/Disadvantages

+     Invariant to image rotation
+     Invariant to intensity changes (partly to multiplicative intensity change)

-     Not invariant to image scaling

K-means Advantages/Disadvantages

+     Simple to implement (unsupervised)
+     It is parallelizable
+     Works well for a large range of problems without need of tuning

-     Variations in cluster size
-     Variations in cluster densities
-     Non-globular shapes
-     Number of clusters need to be decided beforehand: inappropriate choice will give poor clustering results

K-nn Advantages/Disadvantages

+     Very flexible model
+     Easy to integrate new images
+     Easy to integrate new classes

-      But expensive at test time 
-      K has to be chosen beforehand
-      Incorrect classifier when the class distribution is skewed.
     +     Weight the classification

Panorama Advantages/Disadvantages

- Reliable Detector: Detect the same point independently in both images
- Reliable & Destinctive descriptor: For each point correctly recognize the corresponding one 

SIFT Advantages/Disadvantages

+     Stability
+     Descriptiveness

Robust matching technique (to noise):
+     Scale invariant
+     Rotation invariant
+     Intensity change invariant

-     Not invariant to light color change

-     Too big when storing each SIFT (128 dimensions), we better can use K-means Clustering to save a lot of space.

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
Remember faster, study better. Scientifically proven.
Trustpilot Logo