近年來,由於數位時代的來臨,大量人類文化產物均被數位化以作呈現和存取便利所用。然而,由於數位化內容是較脆弱且容易損失的,因此探討如何能夠長久保存數位化的人類文化遺產之數位典藏議題,已成為各國家大型典藏組織所最關切之事。因此針對數位影像典藏之議題,我們為解決現今國際間各典藏系統所實作的轉置和模擬保存策略之缺點,在本論文中我們提出一個具有彈性組態架構之影像轉碼器,以期使典藏系統可彈性建構各影像轉碼器來完成轉置保存的動作,並同時有效降低系統不必要的管理負載。此外,由於影像轉置的品質會因不同的影像內涵而有所變化,因此為使轉置效果最佳化,我們也於典藏系統中設計一智慧影像圖文分類系統。透過此系統,我們可分析典藏影像之內涵而判別出其為圖型或文字類型,此將可提供使用者轉置到最適當影像格式之參考。從實作成果中可證明,本論文所提出的數位影像典藏系統除了可有效降低維謢系統的負載,對於轉置最佳化所需的影像分類功能也具有良好的分類效果以提供轉置的影像格式之選擇。
As the digital era is coming, enormous human culture heritages, such as historic photographs, artistic creations or archival ancient objects, are being digitized for archiving. However, digital information is more fragile and sensitive to data loss; therefore, a lot of works have been presented to discuss the digital preservation for keeping digital data accessible after a long time. To deal with the system maintainance problem caused by the popularly adopted preservation strategies of migration and emulation techniques, we develop a flexible digital image transcoder to arbitrarily construct any kind of encoder-decoder pairs for digital format transcoding. Since the transcoded image could strongly depend on source image content and the migrated image format, we propose an intelligent text-graphic image classifier, which analyzes the source image contents and provides information for migrated image format selection. Experimental results show that our digital image preserving system can be achieved by transcoding strategy while managing the transcoding system in an efficient way. Our classifier can also successfully indentify source image content for migrating to a suitable image format.