講演抄録/キーワード |
講演名 |
2005-01-24 15:15
SOMを用いたフラクタル画像符号化 ○中野博喜・安達雅春(東京電機大) |
抄録 |
(和) |
フラクタル画像符号化の一種であるPIFS(Partitioned Iterated Function Systems)は,画像内の自己相似性を用いて符号化する方法である.PIFS は原画像を互いに重なり合わないブロック(Range Block)に分割し,縮小した原画像の中から最も相似なブロック(Domain Block)を探し出す.そして,その対応関係を符号化する.フラクタル画像符号化は高い圧縮率が期待されているが,各Range Block に対して最も相似なDomain Block を探し出すには多くの計算コストを必要とする.そこで,SOM(Self-Organizing Maps:自己組織化マップ)が持つ分類能力を用いてRange Block とDomain Block を分類し,探索にかかる計算コストを削減する.本手法を用いることにより,画質の劣化を抑えつつ大幅な計算コストの削減が可能となることを示す. |
(英) |
Partitioned Iterated Function Systems that are called PIFS, encode an image using a property of fractal in the image. Therefore, PIFS is a kind of fractal image coding. PIFS partitions the original image into some blocks that are called range blocks. The range blocks are assigned so that they do not overlap with each other. For each range block the most similar block, that is called domain block, is searched in the contracted image. The
image is encoded by the correspondences of the range blocks to the domain blocks. Although fractal image oding
is expected the high compression rate, searching the optimal domain block to each range block has been equired
large computational costs. Therefore, we propose a new coding method using Self-Organizing Maps for lassifying
range blocks and domain blocks. By classifying the range blocks and domain blocks the computational costs are decreased. This method can decrease the computational costs in comparison to the original PIFS though it keeps the image quality almost the same with that of the original PIFS. |
キーワード |
(和) |
フラクタル画像符号化 / 自己組織化マップ / / / / / / |
(英) |
Fractal image coding / Self-Organizing maps / / / / / / |
文献情報 |
信学技報, vol. 104, no. 583, NLP2004-93, pp. 27-30, 2005年1月. |
資料番号 |
NLP2004-93 |
発行日 |
2005-01-17 (NLP) |
ISSN |
Print edition: ISSN 0913-5685 |
PDFダウンロード |
|