基於自我運動速度估測分析之數位影像穩定器

Abstract

數位影像穩定器，主要為了移除影片中非預期的運動向量，是應用在手持相機錄影或行車紀錄器上一個理想相機必須擁有的功能。優秀的數位影像穩定器應能濾除抖動運動向量，但仍能保留所欲的運動向量。

在本文中，我們基於多個幀之間運動向量算得的速度來進行分析，並提出一個新的數位影像穩定器。透過更準確的自我運動向量估計演算法，我們處理的影片成果能比其他已提出方法有更加的視覺品質。

The digital image stabilization (DIS), which removes unwanted motions, is a desirable camera capability for taking videos with a hand-held camera or a recorder installed on moving vehicles. A good DIS should be able to remove jitter motions but still remains the intentional ego-motion. In this thesis, we present a new DIS based on motion velocity analysis among several successive frames. By relying on a more accurate ego-motion estimation algorithm, the processed video shows better visual quality than other published works.