The number of octaves is a crucial parameter in the SIFT algorithm. However, according to [SIFT documentation](http://docs.opencv.org/trunk/d5/d3c/classcv_1_1xfeatures2d_1_1SIFT.html) in OpenCV you can set only the number of layers per octave (`nOctaveLayers`) and not the number of octaves used.
So what is the default number of octaves used by SIFT in OpenCV? And how to change that value?
In addition, it's not clear what is an optimal value for the number of layers per octave. According to [this](http://docs.opencv.org/3.1.0/da/df5/tutorial_py_sift_intro.html) article:
> paper gives some empirical data which> can be summarized as, number of> octaves = 4, number of **scale levels = 5**
However, the default value given to `nOctaveLayers` is 3.
↧