• 作者： 陳玲慧; 陳佑冠
• 作者服務機構： 國立交通大學資訊科學研究所
• 中文摘要： 本論文提出一新的影像壓縮法。此法之主要精神在於將一影像分割成若干區塊，並以區塊之邊界點及內部紋理值來描述此影像。據此，本文首先提出一新的分區法。此法先將具相似灰暗值之連續像素點歸為同一區塊，再利用邊界資訊合併同屬一目標物或背景之相鄰區塊，因邊界資訊為人類視覺系統對物體感應之重要特微，故多數目標物及背景可被完整擷取。而後，對每一區塊，本文利用一多項式表示其內部紋理值，並利用週邊編碼法描述其邊界點。實驗結果顯示本法確可達高壓縮倍率，並可保存影像之重要特徵。
• 英文摘要： In this paper, an image compression method is proposed. First, it presents a new segmentationtechnique to divide an image into several regions. Each region is then represented by its interior textureand its contour. The method then uses an individual polynomial function to approximate the texture ofone region and uses the contour encoding technique to describe its boundary points. To reduce the falsecontours which exist in the reconstructed image, the proposed segmentation technique uses the edgeinformation. Experimental results show that the proposed approach can obtain a high compression ratioand also can preserve the main features of an image. Furthermore, in order to make the proposed methodalso suitable for some applications needing high quality, an alternate coding technique for the texture ofeach region is provided. It can obtain quality as high as required, but the compression ratio will be loweredrelatively.
• 中文關鍵字： contour; edge detection; polynomial function; region extractor; texture
• 英文關鍵字： --