1077 |
|
|
1078 |
\subsection{Recognizability and memorizability} |
\subsection{Recognizability and memorizability} |
1079 |
|
|
|
JVK |
|
|
|
|
|
- lots of texture perception work on texture discrimination |
|
1080 |
|
|
1081 |
- we need also memorization; therefore, careful evaluation is needed |
JVK |
|
|
|
|
- our textures are on a higher level, more like complete pictures |
|
1082 |
|
|
1083 |
- experiments on black-and-white (faces,) ink blots, and snow crystals |
There has been |
1084 |
\cite{goldstein71visualrecognition} show that |
lot of texture perception work on texture discrimination. |
1085 |
complex pictures can be remembered and recognized and that recognition |
However, in our application texture discrimination is not as |
1086 |
performance decreases very little over time. |
much of an issue as memorizability and recognizability of |
1087 |
|
previously seen textures. |
1088 |
- pilot test: 15 textures or pure colors shown sequentially, 5 seconds each. |
Furthermore, our textures are on a higher level, |
1089 |
recognition tested by showing the 15 seen textures mixed with 15 new |
more like complete pictures than the usually studied microstructure. |
1090 |
textures in a random order, subject answers old/new. |
|
1091 |
|
Experiments on black-and-white %(faces,) |
1092 |
- results: for textures, 80 percent correct (12/15 correctly recognized |
ink blots, and snow crystals |
1093 |
as seen and 12/15 correctly identified as not seen). |
\cite{goldstein71visualrecognition} show that |
1094 |
for pure colors, 53 percent correct (13/15 correctly recognized |
complex pictures can be remembered and recognized and that recognition |
1095 |
as seen and 3/15 correctly identified as not seen). |
performance decreases very little over time. |
1096 |
|
|
1097 |
- the experiment only measured the ability to dicriminate previosly seen |
We have conducted a pilot experiment (with one subject) |
1098 |
textures from unseen textures. However, our experience shows that |
in a similar setting comparing |
1099 |
at least the most recurring textures can easily be associated to the |
the recognition performance of our textures with pure color backgrounds. |
1100 |
document content. |
First, 15 textures (or pure colors) were shown sequentially, 5 seconds each. |
1101 |
|
Then, recognition was tested by showing the 15 seen textures mixed |
1102 |
|
with 15 unseen textures in a random order, and the subject answered |
1103 |
|
old or new. |
1104 |
|
The results were as follows: |
1105 |
|
for textures, 80 percent correct (12/15 correctly recognized |
1106 |
|
as seen and 12/15 correctly identified as not seen); |
1107 |
|
for pure colors, 53 percent correct (13/15 correctly recognized |
1108 |
|
as seen and 3/15 correctly identified as not seen). |
1109 |
|
|
1110 |
|
Of course, this result is not statistically relevant, |
1111 |
|
but in conjunction with our experiences, it suggests that the |
1112 |
|
textures are quite recongizable and that they are clearly |
1113 |
|
superior to pure colors in terms of recognizability. |
1114 |
|
|
1115 |
|
Furthermore, the experiment only measured the ability to |
1116 |
|
recognize previosly seen textures within unseen textures. |
1117 |
|
Our experience shows that in practice, |
1118 |
|
at least the most recurring textures are not only recognized but can |
1119 |
|
also easily be associated to the document content. |
1120 |
|
|
1121 |
- as per Zipf's law [XXXref], it suffices to learn the textures |
- as per Zipf's law [XXXref], it suffices to learn the textures |
1122 |
of the most recurring documents. |
of the most recurring documents. |