1306 |
} |
} |
1307 |
\end{figure} |
\end{figure} |
1308 |
|
|
1309 |
\subsection{Recognizability experiment} |
\subsection{An Initial Recognizability Experiment} |
1310 |
|
|
1311 |
%In our application visual texture discrimination is not as |
%In our application visual texture discrimination is not as |
1312 |
%much of an issue as memorizability and recognizability of |
%much of an issue as memorizability and recognizability of |
1326 |
|
|
1327 |
In order to evaluate the recognizability of our procedurally generated |
In order to evaluate the recognizability of our procedurally generated |
1328 |
textures, we need to have an appropriate comparison point. |
textures, we need to have an appropriate comparison point. |
1329 |
Pictures of natural objects would not provide a good comparsion, |
Pictures of natural objects would not be appropriate, |
1330 |
because they cannot be generated in infinite amounts from seed |
because they cannot be generated in infinite amounts from seed |
1331 |
numbers and they would esasily yield undesirable semantic associations. |
numbers and they would esasily yield undesirable semantic associations. |
1332 |
Thus, we simply use plain solid color backgrounds as a baseline for comparison |
Lacking a better example, we shall use plain solid color backgrounds |
1333 |
|
as a baseline |
1334 |
even though the colors of even a small set of randomly chosen colors would |
even though the colors of even a small set of randomly chosen colors would |
1335 |
most likely not be discriminable. |
most likely not be discriminable. |
1336 |
|
|
1337 |
Surely it would be hard to remember a large number of textures, too. |
Another question is how many textures would the user have to remember |
1338 |
However, the user does not have to remember all the textures; |
for it to be useful. |
1339 |
it suffices to learn the textures of the most often used documents |
Studies of web cache statistics (see, e.g. \cite{breslau99web}) |
1340 |
and distinguish them from the rest: |
have shown that file popularity approximately follows Zipf's law |
1341 |
|
so that a small number of documents accounts |
1342 |
|
for most of the use, |
1343 |
|
%Obviously, the user does not have to remember all the textures; |
1344 |
|
%it suffices to learn the textures of the most often used documents |
1345 |
|
%and distinguish them from the rest, |
1346 |
see Fig.~\ref{fig-zipf}. |
see Fig.~\ref{fig-zipf}. |
1347 |
Many studies of web cache statistics |
|
1348 |
have shown that for instance file popularity approximately follows Zipf's law |
|
|
(see, e.g. \cite{breslau99web}) so that a small number of documents accounts |
|
|
for most of the use. |
|
1349 |
%Also \cite{glassman94caching}, perhaps the first web cache zipf ref |
%Also \cite{glassman94caching}, perhaps the first web cache zipf ref |
1350 |
|
|
1351 |
\begin{figure} |
\begin{figure} |
1363 |
Each square represents a document, and the area of each square is scaled |
Each square represents a document, and the area of each square is scaled |
1364 |
to its rate of accesses. |
to its rate of accesses. |
1365 |
The diagram shows 2000 documents weighted with Zipf's law with exponent 1.1. |
The diagram shows 2000 documents weighted with Zipf's law with exponent 1.1. |
1366 |
Here, the 15 most important documents account for approximately half of the accesses. |
Under these conditions, the 15 most important documents account for approximately half of the accesses. |
1367 |
Of course, it is impossible to know which documents will be important and |
Of course, it is impossible to know which documents will be important and |
1368 |
that will also change with time so all documents should be textured |
that will also change with time so all documents should be textured |
1369 |
stably from the first viewing onwards. |
stably from the first viewing onwards. |