Description
SIFT (Scale Invariant Feature Transform)
SIFT is a well-known algorithm in computer vision to detect and describe local features in
images. Its applications include object recognition, robotic mapping and navigation, image
stitching, 3D modelling, video tracking and others.
A SIFT feature is a salient keypoint that corresponds to an image region and has an
associated descriptor. SIFT computation is commonly divided into two steps:
• detection
• description
At the end of the detection step, and for each feature detected, the SIFT algorithm
establishes:
• keypoint spatial coordinates (x, y)
• keypoint scale
• keypoint dominant orientation
After the detection step, the description step computes a distinctive fingerprint of 128
dimensions for each feature. The description obtained is designed to be invariant to scale
and rotation. Moreover, the algorithm offers decent robustness to noise, illumination
gradients and affine transformations.
Lab Tasks
Task 1: Compute SIFT features:
• Extract SIFT features with default parameters and show the keypoints on the image
• Check (i.e. print) the SIFT extraction results: How many descriptors are extracted?
What are the descriptor values?
• Reduce the number of keypoints extracted so that the visualisation of keypoints is
easier. Show these keypoints on the image. (Hint: vary the parameter
“contrastThreshold” or “nfeatures” so that the number of keypoints becomes about
¼ of all default keypoints).
Task 2: Rotate the image and compute the SIFT features again:
• Rotate the image clockwise by 45 degrees
• Extract SIFT features and show the keypoints on the rotated image, using the same
parameter settings as task 1 (for reduced number of keypoints)
• Inspect the keypoints visually: Do the keypoints look roughly the same as those
extracted on the original image? What does this observation imply?
Submit the two images showing the keypoints produced in tasks 1 and 2 for marking. Note
that the image should be converted to greyscale for SIFT computation, but keypoints can be
shown on colour images to have a better visualisation.
NOTE: Refer to https://docs.opencv.org/3.4.3/da/df5/tutorial_py_sift_intro.html for an
example of computing SIFT features and showing the keypoints. The attached template.py
file can be used as a template for the lab tasks.
REFERENCES
D. G. Lowe, “Distinctive image features from scale-invariant keypoints,” Int. J. Comput. Vis., vol. 60,
no. 2, pp. 91–110, 2004
