| DC 欄位 |
值 |
語言 |
| DC.contributor | 資訊工程學系 | zh_TW |
| DC.creator | 黃韻文 | zh_TW |
| DC.creator | Yun-Wen Huang | en_US |
| dc.date.accessioned | 2017-7-13T07:39:07Z | |
| dc.date.available | 2017-7-13T07:39:07Z | |
| dc.date.issued | 2017 | |
| dc.identifier.uri | http://ir.lib.ncu.edu.tw:444/thesis/view_etd.asp?URN=104522058 | |
| dc.contributor.department | 資訊工程學系 | zh_TW |
| DC.description | 國立中央大學 | zh_TW |
| DC.description | National Central University | en_US |
| dc.description.abstract | 當人們觀賞一張照片時，會常常想像在照片中的影像如果動起來會是什麼樣，想像力讓我們大腦接收到的資訊超乎一張靜態影像能給予我們的，在電腦視覺這門科學中，如何在靜態影像中加入動態的畫面的研究也正持續進行著。我們提出了一個能夠合成水流動畫到單張影像的系統，讓編輯者選擇想要編輯的影像區域和來源影片。我們分析水流的方法整合了線卷積與光流法，將影像與影片轉換成具有水流流向特徵的紋理貼圖，藉由分析此紋理貼圖來比對水流的相似程度，可以減少影像中許多容易影響比對的因素，像是顏色、明暗等等的。在做合成之前將來源影片的顏色外觀轉換成與目標影像相似，可以減少找不到適合的接縫的機率，本篇論文最後實作的紋理合成的演算法是基於影像縫合與圖切割的概念來完成，最後在套上多重解析度影像融合在人為痕跡明顯的拼接處，將不連續感進一步消除。 | zh_TW |
| dc.description.abstract | When people view a picture, they often imagine how the dynamic object in picture would move. The imagination makes people perceive much more than a static object before them and adding the motion in static image is always an active area in computer graphics. This paper present a system for synthesizing fluid motion on a single image. We include multiple relative techniques of image processing and computer vision.
The flow animation is extracted from a related video sequence and pasted onto the target image. For reducing the complexity of our system, we use a semi-automatic approach to let the user control the target image and the source video to create desired result. We build a simple interface for the following user-defined process. The user manually specifies the interested region of target image along with segmentation by grab-cut and choose the reference video. The user semi-automatically set the trimap for alpha matting to extract the foreground of video. Then, the system automatically executes remain process.
Our method integrates the optical flow and the line integral convolution to transfer the target image and video to textures that has correlation of pixels along the flow. The flow field of video can be obtained by compare the consecutive frame but the for single image there has no pervious frame can be analysis. We simply calculate the gradient of segmented target to generate the approximate flow field. The fluid orientation between image and video can be compared without the color’s influence by using the texture-based flow image.
The synthesis algorithm we purposed in this paper is based on the idea of image quilting and the graph-cut. The matching process depends on the flow texture but the synthesis process uses the original texture. The weight computation of seam finding on each node is refer to the color difference. Before synthesizing, we transfer color appearance of the video frame to be like the target image that can decrease the probability for visible seams in the generated result. Finally, we refine the appearance of the result by blending or blurring the seam. | en_US |
| DC.subject | 紋理與視頻合成 | zh_TW |
| DC.subject | 基於圖像的渲染 | zh_TW |
| DC.subject | 圖像與視頻處理 | zh_TW |
| DC.subject | 基於紋理的可視化 | zh_TW |
| DC.subject | 自然場景 | zh_TW |
| DC.subject | Texture and video synthesis | en_US |
| DC.subject | image-based rendering | en_US |
| DC.subject | image and video processing | en_US |
| DC.subject | texture-based flow visualization | en_US |
| DC.subject | natural phenomenon | en_US |
| DC.title | 創建靜態影像中的水流動畫 | zh_TW |
| dc.language.iso | zh-TW | zh-TW |
| DC.title | Creating Fluid Animation in Static Image | en_US |
| DC.type | 博碩士論文 | zh_TW |
| DC.type | thesis | en_US |
| DC.publisher | National Central University | en_US |