發(fā)布時間：2018-04-03 17:17

  本文選題：圖像瀏覽　切入點：照片拼貼　出處：《哈爾濱工業(yè)大學》2013年碩士論文

【摘要】：數(shù)碼照片和視頻數(shù)據(jù)的急劇增長需要既能支持速度快又能支持內(nèi)容形象化瀏覽的表達技術(shù)。隨著圖像文件數(shù)量的爆炸式增長，管理大量圖像的內(nèi)容的能力已成為一項關(guān)鍵技術(shù)。 一種最有效且視覺上有吸引力的圖像瀏覽技術(shù)是基于拼貼的表達。拼貼表示的是一個圖像庫或視頻材料，作為有意義的快照集合，它允許對庫的內(nèi)容進行概述。 自動拼貼設(shè)計在很多應用中，從家庭拼貼的設(shè)計到商業(yè)廣告項目，都是非常普遍的任務(wù)。 在我們的研究中，我們提出了兩種自動拼貼圖組成方法：基于ROI形狀估計的方法和基于ROI自適應塊分析的方法。 首先提出的基于ROI形狀估計的方法與國際現(xiàn)有技術(shù)具有相近的組成性能、信息和美學價值。 改進的基于ROI自適應塊分析的方法比現(xiàn)有技術(shù)效果好。根據(jù)一系列實驗分析和基于社會網(wǎng)絡(luò)的調(diào)查（多達300個受訪者進行評估），，完成拼貼圖組成的整個過程（包括幀選擇、分配和ROI近似）的平均時間為17.98秒。生成布局的多樣性為81.16%，伴隨著空區(qū)域為穩(wěn)定的0%級；ROI近似的準確率為86.08%；整個輸出的拼貼的信息價值和美學價值分別估計為90.25%和90.83%。 基于ROI自適應塊分析的方法比Autocollage大約快4秒，并且輸出的拼貼比Autocollage高出5%的信息價值和10%的美學價值。 我們提出的方法既可以用于管理圖像庫，也可以用于管理視頻材料，并且能夠編輯更特殊的輸出拼貼。
[Abstract]:The rapid growth of digital photo and video data requires representation techniques that can support both fast and visualized browsing of content.With the explosive growth of the number of image files, the ability to manage the content of a large number of images has become a key technology.One of the most effective and visually attractive image browsing techniques is collage based expression.Collage represents an image library or video material that allows an overview of the contents of the library as a meaningful collection of snapshots.Automatic collage design is a common task in many applications, from home collage design to commercial advertising projects.In our research, we propose two automatic mapping methods: one based on ROI shape estimation and the other based on ROI adaptive block analysis.First, the proposed method based on ROI shape estimation has similar composition performance, information and aesthetic value to the existing international technology.The improved method based on ROI adaptive block analysis is more effective than the existing technology.According to a series of experimental analyses and social network-based surveys (up to 300 respondents were evaluated, the average time to complete the whole process of mapping (including frame selection, allocation and ROI approximation) was 17.98 seconds.The diversity of the generated layout is 81.16, and the accuracy of the approximation is 86.08 with a stable 0% ROI of empty area, and the information value and aesthetic value of the whole output collage are estimated to be 90.25% and 90.833%, respectively.The adaptive block analysis method based on ROI is about 4 seconds faster than Autocollage, and the output collage is 5% higher information value and 10% higher aesthetic value than Autocollage.The proposed method can be used to manage both image libraries and video materials, and to edit more special output collages.
【學位授予單位】：哈爾濱工業(yè)大學
【學位級別】：碩士
【學位授予年份】：2013
【分類號】：TP391.41

【共引文獻】

相關(guān)碩士學位論文 前1條

1 岑磊;基于個性化推薦的圖像瀏覽與檢索相關(guān)方法研究[D];復旦大學;2011年

本文編號：1706152