D = VL_LIOP(I) computes the LIOP descriptor of an image I, as described by [1]. I is a gray-scale square image with odd side length of class SINGLE. D is a column vector containing the LIOP descriptor of I. Note that LIOP is also integrated in the VL_COVDET() function for feature extraction.
VL_LIOP() accepts the following options:
- NumNeighbours 4
Set the number of neighbours sampled to construct the order pattern of each image pixel.
- Radius 5
Set the radius of the circular neighbourhood used to sampled the local order pattern of each pixel.
- NumSpatialBins 6
Set the number of spatial pooling regions. The LIOP descriptor has dimension factorial(NumNeighbours) * NumSpatialBins.
- IntensityThreshold -0.02
Set the intensity threshold used to weight order patterns as they are pooled into a histogram. A negative value is interpreted as a fraction of the difference between the maximum and minimum intensity in each local patch.
- Verbose
If specified, be verbose
- REFERENCES
[1] Z. Wang, B. Fan, F. Wu. Local Intensity Order Pattern for feature description. In ICCV, 2011