55 |
|
|
56 |
\begin{abstract} |
\begin{abstract} |
57 |
|
|
|
%FIXME: more references of texture generation |
|
|
|
|
58 |
%FIXME: Justify why adding unique background will help recognise a document |
%FIXME: Justify why adding unique background will help recognise a document |
59 |
%from a fragment and enhance text readability. (Optional) |
%from a fragment and enhance text readability. (Optional) |
|
|
|
|
%FIXME: Add a concise description of how fig. 3 works in the paper. |
|
|
|
|
|
%FIXME: Make the titles of figures shorter, and explain the details in the |
|
|
%context.(Optional) |
|
|
|
|
|
%FIXME: Revise the spelling and grammar errors, esp. grammar of long sentences. |
|
|
|
|
|
|
|
60 |
%FIXME: more concise and exact explanations in some sections |
%FIXME: more concise and exact explanations in some sections |
61 |
%FIXME: what are this paper's specific contributions? (e.g., in Sec. 4) |
%FIXME: what are this paper's specific contributions? (e.g., in Sec. 4) |
62 |
|
|
349 |
%of visual perception |
%of visual perception |
350 |
%are fairly well known |
%are fairly well known |
351 |
in the visual cortex, there are cells sensitive to different |
in the visual cortex, there are cells sensitive to different |
352 |
frequencies, orientations, and locations in the visual field%(see, e.g.,~Bruce et al |
frequencies, orientations, and locations in the visual field%(see, e.g., Bruce et al |
353 |
\cite{bruce96visualperception}. |
\cite{bruce96visualperception}. |
354 |
|
|
355 |
On a higher level, the correlations between local features are combined |
On a higher level, the correlations between local features are combined |
491 |
% |
% |
492 |
% For instance, consider a published scientific article |
% For instance, consider a published scientific article |
493 |
% available through several different URLs on the WWW, |
% available through several different URLs on the WWW, |
494 |
% e.g.,~through ResearchIndex\cite{giles98citeseer%,lawrence99digital |
% e.g., through ResearchIndex\cite{giles98citeseer%,lawrence99digital |
495 |
% }, |
% }, |
496 |
% the ACM portal and the author's web site. |
% the ACM portal and the author's web site. |
497 |
% Now, when a user downloads and saves one copy, the relationship between |
% Now, when a user downloads and saves one copy, the relationship between |
615 |
a limited |
a limited |
616 |
number of different features detected can be grouped into |
number of different features detected can be grouped into |
617 |
objects, indicating that the spatial resolution of the feature |
objects, indicating that the spatial resolution of the feature |
618 |
vector is quite low -- as a well-known example, conjunction |
vector is quite low --- as a well-known example, conjunction |
619 |
coding is not preattentive -- red squares are hard to find among |
coding is not preattentive --- red squares are hard to find among |
620 |
green squares and red and green circles. |
green squares and red and green circles. |
621 |
|
|
622 |
The feature vector is then used to compute |
The feature vector is then used to compute |
640 |
The qualitative model of visual perception used to create |
The qualitative model of visual perception used to create |
641 |
the algorithm. |
the algorithm. |
642 |
The visual input is transformed into a feature vector, |
The visual input is transformed into a feature vector, |
643 |
which contains numbers (activation levels) corresponding to |
which contains numbers (activation levels) corresponding to, |
644 |
e.g.~colors, edges, curves and small patterns. |
e.g., colors, edges, curves and small patterns. |
645 |
The feature vector is matched against the memorized textures. |
The feature vector is matched against the memorized textures. |
646 |
In order to generate recognizable textures, random seed |
In order to generate recognizable textures, random seed |
647 |
values should produce |
values should produce |
729 |
in order to produce a unique background from |
in order to produce a unique background from |
730 |
a random vector. |
a random vector. |
731 |
Features that are orthogonal for human perception |
Features that are orthogonal for human perception |
732 |
(e.g.,~color and the direction of the fastest luminance change) |
(e.g., color and the direction of the fastest luminance change) |
733 |
should be independently random, and features not orthogonal |
should be independently random, and features not orthogonal |
734 |
(e.g., colors of neighbouring pixels) |
(e.g., colors of neighbouring pixels) |
735 |
should be correlated so as to maximize the entropy. |
should be correlated so as to maximize the entropy. |
1561 |
magazines and web sites have long used skillfully designed |
magazines and web sites have long used skillfully designed |
1562 |
graphical elements to make themselves recognizable. |
graphical elements to make themselves recognizable. |
1563 |
However, our algorithm is able to generate an unlimited |
However, our algorithm is able to generate an unlimited |
1564 |
amount of unique backgrounds cheaply, making it possible |
amount of unique backgrounds cheaply, making it possible, |
1565 |
e.g. for all academic articles with similar typography |
e.g., for all academic articles with similar typography |
1566 |
to have their own background. |
to have their own background. |
1567 |
|
|
1568 |
%\subsection{Further work} |
%\subsection{Further work} |
1599 |
% to use repeating units for non-repeating backgrounds --- but use more than one, |
% to use repeating units for non-repeating backgrounds --- but use more than one, |
1600 |
% which are not rationally related. |
% which are not rationally related. |
1601 |
|
|
1602 |
We report an initial recognition experiment showing that the textures |
We have conducted an initial recognition experiment |
1603 |
|
showing that the textures |
1604 |
generated by our algorithm are indeed recognizable. |
generated by our algorithm are indeed recognizable. |
1605 |
Carrying out usability more tests is necessary, both |
Carrying out more usability tests is necessary, both |
1606 |
to measure how well textures can be remembered |
to measure how well textures can be remembered |
1607 |
and to make the ad hoc distributions more experimentally based. |
and to make the ad hoc distributions more experimentally based. |
1608 |
|
|