Harris-like Scale Invariant Feature Detector (HLSIFD)

Yinan Yu, Kaiqi Huang, Tieniu Tan

For scale invariant local feature modeling, the primitive model is the Gaussian model. It can be described by the second-derivation of the image. Based on the Gaussian model, Harris-like Scale Invariant Feature Detector (HLSIFD) extract more stable local features. This method is based on SIFT and inherit the advantage of Harris-like function.
  • Different to SIFT, HLSIFD does not need a post-treatment step to cut edge-like points suddenly which would affect the stability.
  • Different to Harris corner, HLSIFD is a scale invarint feature detector. HLSIFD can suppress the unstable fake feature points uniformly. Therefore, with fewer meaningless key points, features are more significant.

    I'm working on an implmention of the HLSIFD algorithm under VLFEAT framework for easy using.

    Local Feature Model SIFT points HLSIFD points
    The differences among SIFT (DoG/LoG), SURF (DoH) and HLSIFD
    The SIFT Key points
    The HLSIFD Key points

    This paper can be download here.

    Major References

    [1] Lowe, D.G, "Distinctive image features from scale-invariant keypoints". Int. J. Comput. Vision 60(2) (2004) 91–110
    [2] Bay, H., Ess, A., Tuytelaars, T., Gool, L.V. "Speeded-up robust features (surf)". Comput. Vis. Image Underst. 110(3) (2008) 346–359
    [3] Harris, C., Stephens, M. "A combined corner and edge detection". (1988) 147–151
    [4] Lindeberg, T. "Scale-space theory in computer vision". (1994)
    [5] Mikolajczyk, K., Schmid, C. "Scale & affine invariant interest point detectors". Int. J. Comput. Vision 60(1) (2004) 63–86