一個關於不使用影片分割遮罩的場景背景建立技術與快速場景背景建立之研究

Abstract

MPEG-4採用了場景背景編碼技術，藉由將影片中的背景合併為單一場景背景，此 技術能夠大幅提升背景的編碼效率。在MPEG-4 中，場景產生器利用畫面的簡單平均產 生背景，此方法導致畫面中曾經有前景物體通過的地區產生模糊現象。為了消除此一模 糊現象，MPEG-4建議利用一個背景/前景物體的分割遮罩，讓背景產生器能分辨物體與 背景的點。然而，自動影像分割是相當困難的。切割失誤的區域將導致產生的場景背景 中存在如鬼影一般的物體殘影。為了解決這個問題，我們將在計畫的第一年，提出一個 不需要背景/前景物體分割遮罩的影像混合演算法。該演算法預計利用畫面中背景應比前 景物體較常出現的特性，利用一個計數方法使得只有背景點被混合進場景背景中。 在第二年的計畫中，我們則預計提出一個快速的多重場景背景切割技術。由於場景 背景建立技術中使用了幾何轉換，因此所產生的場景背景會遭受到幾何變形的影響。幾 何變形使得場景背景的實際可用視角因而減小，編碼效率也大幅滑落。為了解決此一問 題，Farin 等人提出了多重場景背景技術，藉由使用多張場景背景，提高可用視角並減 少幾何變形量，Farin 同時亦提出一最佳化的多重場景背景分段技術。然而，此最佳化 演算法需要大量的運算量以尋找出分段的最佳解。為了解決這個問題，我們預計提出一 個快速近似最佳化多重場景背景影像分割方法。我們的快速演算法預計利用影片的全域 位移以及全域放大/縮小資訊，快速得到可能的分段點資訊。並利用所得到的分段資訊， 將影片背景建立為多重場景背景。預期建立出的場景品質及大小將近似於Farin 等人所 提出之方法產生的場景。

Sprite coding, which can increase the coding efficiency of backgrounds greatly, is a novel technology adopted in MPEG-4 object-based coding. The sprite generator introduced in MPEG-4 blends frames by averaging, this will make some places, which are ever occupied by moving objects, look blurring. Thus, providing segmented masks for moving objects is suggested. Since perfect segmentation is impossible, some ghost-like shadows will appear in the generated sprite. To treat this problem, in the first year of this project, an blending strategy without needing segmentation masks will be developed. It is based on the fact that for each point in the generated sprite, the corresponding pixels in most frames belong to background and only few belong to moving objects. A counting schema is provided to make only background points attend in average blending. In the second year of this project, a fast multiple-sprite partition algorithm will be proposed. Due to the geometric transformation applied to each non-reference frame in the procedure of sprite coding, the generated sprite is distorted and the available view angles relative to the reference frame are restricted. These problems can be resolved by using multiple sprites. An optimal multi-sprite generator has been proposed by Farin et. al., but the optimal method requires huge computation in the sprite coding costs. To treat this disadvantage, a fast near-optimal multi-sprite partition algorithm will be developed in this project based on frame translation and scaling information. Our fast algorithm will divide a video sequence into several subsequences. Sprites will be generated from every subsequence, and the total quality and size of all generated sprites will be near those of sprites generated by Farin’s optimal generator.