Image feature detection based on a spider web model

find image features

This folder contains an illustrative example for the implementation of the Spider Local Image Feature (SLIF) descriptor.
The included folder contains the following files:
SLIFMatchingTest.m
extractSLIFFeatures.m
Painting.jpg
Painting(Scale=0.50).jpg
The file “SLIFMatchingTest.m” provides a feature matching example. In such example, a set of SURF features are first detected by applying the Fast Hessian Feature detector, and then, the SLIF description approach (implemented though the function “extracSLIFFeatures.m” that is included in this folder) is applied to assign a descriptor to each of such detected features.

In the SLIF description approach, the orb-web structures used on for the description of feature points are defined by its respective number of radial threads (‘M’) and spiral threads (‘N’). Also, a scaling factor (‘K’) is added to the increase the area occupied by such structures, allowing them to consider bigger pixel neighborhoods in the description process. This parameters may be changed in the following lines of code:

%% SLIF Parameters
% Number or Radial Threads "M"
M = 12;
% Number of Spiral Threads "N"
N = 6;
% Web Scaling Factor "K"
K = 10;

In the SLIF approach, the dimension of the resulting feature descriptors is equal to the product between the numbers of radial threads and spiral threads (M×N). The user is free to change the values to produce different sized descriptors.

Furthermore, the next lines of codes are used to set an image pair “I1” and “I2” from which we want extract and match its features:

%% 1. Read Image Pair.
I1 = imread('Painting.jpg');
I2 = imread('Painting(Scale=0.50).jpg');

The images “Painting.jpg” and “Painting(Scale=0.50).jpg” are both included in this folder for illustrative purposes. The user is free to try different images “I1” and “I2” to test this feature matching example.

For more information please refer to
Fernando Fausto, Erik Cuevas, Adrián Gonzales, A new descriptor for image matching based on bionic principles, Pattern Analysis and Applications, In press.

https://link.springer.com/article/10.1007/s10044-017-0605-z