| Summary: | 碩士 === 國立臺灣科技大學 === 資訊工程研究所 === 88 === Symmetry detection plays an important role in image processing and pattern recognition.
Based on the gradient orientation histogram of the grey image, this thesis proposes a
coarse-to-fine approach for reflectional symmetry detection. Initially, we use part of the
gradient information in the histogram, i.e. we make the quantization interval of the
histogram coarser in order to find out an approximate orientation of the symmetric axis.
This would save the computing time. Then we narrow the searching area of gradient
orientation information in the histogram; meanwhile, we use finer quantization interval
in the searching area. We continue the above procedure until the finest quantization interval
is reached. Ultimately, the orientation of the symmetric axis is obtained. We report the line
passing through the centroid of the object with the orientation obtained as the reflectional
symmetry. Our method only needs the original grey image and the gradient information
of the image. Some experiments have been carried out to demonstrate the computational
advantage of the proposed algorithm when compared to the previous algorithm.
|
|---|
