Feature point description
By the steps mentioned above, there are three information, including position, scale and direction, for each keypoint.
Then a descriptor is built for each feature point,
which is described using a descriptor vector that has 4×4×8=128 dimensions.
This kind of descriptor vector has little change with image translation such as scale and illumination.
This descriptor not only includes the feature points, but also contains the contribution of its surrounding pixels.
In addition, the descriptor should have high uniqueness so as to improve the matching accuracy of feature point