[Gzz-commits] journals/plans SA_2003.rst

[Janne V. Kujala]

CVSROOT:        /cvsroot/fenfire
Module name:    journals
Changes by:     Janne V. Kujala <address@hidden>        03/05/14 08:12:04

Modified files:
        plans          : SA_2003.rst 

Log message:
        text filtering

CVSWeb URLs:
http://savannah.gnu.org/cgi-bin/viewcvs/fenfire/journals/plans/SA_2003.rst.diff?tr1=1.136&tr2=1.137&r1=text&r2=text

Patches:
Index: journals/plans/SA_2003.rst
diff -u journals/plans/SA_2003.rst:1.136 journals/plans/SA_2003.rst:1.137
--- journals/plans/SA_2003.rst:1.136    Wed May 14 08:07:57 2003
+++ journals/plans/SA_2003.rst  Wed May 14 08:12:04 2003
@@ -494,8 +494,6 @@
 Text filtering
 ''''''''''''''
 
-JVK
-
 Most user interfaces only use pre-rendered character
 bitmaps of the font, which forces text to be 
 drawn on straight horizontal lines with constant size
@@ -516,17 +514,21 @@
 possible to run small *fragment programs* for each rendered pixel.
 This flexibility allows us to overcome the restrictions in the
 texture filtering.
-
 However, designing such filters in an *ad hoc* fashion is 
-extremely cumbersome and error-prone. What is needed is a suitable
-mathematical framework for modeling the
-*perceptual* qualities of rendered text.
-
-The goals in this area are the development of the mathematical framework
-for qualitatively 
-XXX
-As one theoretical approach for modeling the readability 
-we can use the correlation function of the rendered text.
+extremely cumbersome and error-prone; a more
+formal approach for evaluating the results of 
+different filtering methods is needed.
+
+The goals in this area are the development of a mathematical framework
+for modeling the *perceptual* qualities of rendered text
+as well as experimental evaluation of the models.
+
+As one theoretical approach for the readability modeling 
+we plan to use the correlation function of the rendered text.
+There have not been much research on this approach in 
+graphics context.
+However, we hope to be able to apply our earlier experience 
+on Bayesian methods and similar models.
 
 
 Unique background textures