this is for holding javascript data
Volker Strobel added section_Analysis_Determining_the_Number__.tex
almost 8 years ago
Commit id: 2787e6f3f76ac62b1126afb8030063846feaebd7
deletions | additions
diff --git a/section_Analysis_Determining_the_Number__.tex b/section_Analysis_Determining_the_Number__.tex
new file mode 100644
index 0000000..2092b1d
--- /dev/null
+++ b/section_Analysis_Determining_the_Number__.tex
...
\section{Analysis -- Determining the Number of Textons}
\label{sec:numtextons}
The developed framework allows to tune the computational complexity by
modifying the number of extracted image patches (samples). To increase
the speed of the algorithm, the goal is to use as few samples as
possible. To determine a suitable number of extracted samples, in this
experiment, the average cosine similarity between $D = 20$ datasets of
histograms is compared. Each dataset consists of $N = 10300$
histograms. The independent variable is the number of extracted image
patches $M$. The histograms were generated using the same images. Due
to the random sampling of the extracted image patches, the histograms
of each datasets will differ. This deviation will be measured using
the cosine similarity. Therefore, each of the $D$ datasets was
compared to all the other $D - 10$ datasets and the average cosine
similarity was determined as well as the standard deviation of the
cosine similarity was measured. Comparing the cosine similarity
between the histograms has the advantage that the number of samples
can be determined independent of a specific task.