CHINESE CALLIGRAPHY IMAGE ANALYSIS PROJECT
Model-Based Analysis of Chinese
Calligraphy Images for Virtual Brush Synthesis
|
|
|
|
CONTENT
TEAM MEMBERS
- Prof.
IP Ho-Shing., Horace (cship@cityu.edu.hk)
- Dr LEUNG Wing-Ho, Howard
(howard@cityu.edu.hk)
- WONG Tak-Sum,
Sam (tswong@cityu.edu.hk)
INTRODUCTION
In the past, Chinese fonts with smooth outlines and solid colouring have been
produced for using in computer displays and printed hard-copies. However,
aesthetic properties of the characters produced by calligraphers could not be
simulated at all.
Nowadays, with the advance of technology, Chinese calligraphic artwork can be
scanned and stored in the form of images. However, the visual quality degrades
a lot when different transformations such as sheering, scaling and rotating are
applied to the characters. As a result, vector-based representation is a better
choice. The objective of this project is to obtain a vector-based
representation from the analysis of Chinese calligraphy images.
PRIOR WORK
|
|
|
|
In the Virtual
Brush project, Prof.
IP Ho-Shing, Horace and Dr WONG Tsui-Fong,
Helena proposed a novel model-based approach to synthesize realistic Chinese
calligraphic writings for generating hair-brush writings such that the
rendering is scalable in resolution and allows high quality publishing. While
this graphical model facilitates the synthesis of calligraphy writing given a
set of writing parameters, for the inverse problem, a lot of user stroke
manipulation is required to regenerate the model parameters given a
calligraphic image. Consequently, an intelligent method is required to
automatically determine the model parameters from images of Chinese
calligraphy.
OBJECTIVES
- Analyze the
Chinese calligraphic images and extract the underlying strokes
- Obtain a
vector-based representation of Chinese calligraphy image
- Estimate the
set of 3D geometric as well as the dynamic writing parameters of the Virtual
Brush model automatically from images of Chinese calligraphic
artwork
PROJECT DESCRIPTIONS
- Images scanned
from real calligraphic artwork are stored first. The ink is then extracted
from the background paper.
|
|
|
|
- Characters are
separated out while for each character the strokes are segmented so that
it can be modelled by several strokes with different parameters.
|
|
|
|
- Each stroke is
then fitted with a set of ellipses with different geometric information
such as size and orientation.
|
|
|
|
|
|
|
|
|
|
- The parameters
of the bristle hairy brush such as the radius of brush’s stem is estimated
according to the features of the strokes:
|
|
|
Radius
of Brush Stem |
|
|
|
|
- The ink
depositing level of each stroke along the trajectory is then estimated.
|
|
|
|
|
|
|
Original Stroke’s Image |
|
Result of Ink Depositing Level’s Estimation |
|
Synthesised Stroke’s Image |
|
|
|
|
|
|
|
Original Stroke’s Image |
|
Result of Ink Depositing Level’s Estimation |
|
Synthesised Stroke’s Image |
-
All the
estimated stroke images forming a character are combined to synthesis
the whole character.
|
|
|
|
|
Estimated Strokes of the
Whole Character Image |
|
Original image in Seal
Character ( |
In
addition to the parameters mentioned above, we also need to estimate the other
parameters in the brush geometric model, brush motion dynamic model and the ink
depositing model:
- The vertical
distance from the root of hairs bundle to the paper;
- The spreading
factor which models various brush movements;
- The absorption
variant which is used to model the time that the brush spent in contact
with a point on the paper;
- The properties
of writing paper which is used to model the different properties such as
the ink spreading rate of the writing paper; and
- The maximum ink
spreading variation degree which is used to model the deviation of each
individual piece of hair.
RESULT
The following images show the results in terms of the synthesised character images using the parameters estimated from real Chinese calligraphy images with our proposed algorithm.
|
|
|
|
Original image in Clerical
Style (隸書) |
Synthesised image using our
estimated model parameters |
|
|
|
|
Original image in Cursive style ( |
Synthesised image using our
estimated model parameters |
)PUBLICATION
Ÿ Sam T.S. Wong, Howard Leung and Horace H.S. Ip, "Automatic Estimation of Ink Depositing Model Parameters from Analysis of Chinese Calligraphic Images", Pacific Graphics 2006 (PG 2006), Taipei, Taiwan, October 2006.
Ÿ Sam T.S. Wong, Howard Leung and Horace H.S. Ip, "Fitting Ellipses to a Region with Application in Calligraphic Stroke Reconstruction", International Conference on Image Processing (ICIP 2006), Atlanta, USA, October 2006.
Ÿ
Sam T.S. Wong, Howard Leung and Horace H.S. Ip,
"Brush Writing Style Classification from Individual Chinese Characters",
International Conference on Pattern Recognition (ICPR 2006), Hong Kong,
August 2006.
Ÿ
Sam T.S. Wong, Howard Leung and Horace H.S. Ip, "Model-based Analysis of Chinese Calligraphy
Images", 9th Intl. Conf. on Information Visualisation (IV05), London,
England, July 2005.
CONTACT
Any suggestions or comments are welcome. Please
send them to Dr Howard LEUNG.
ACKNOWLEDGEMENTS
We
thank the Chinese Civilisation
Centre for allowing us to use the wall-paper in this webpage.