http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Video Stabilization using Clustered Directional Features
Thien-Pham Minh(팜민티엔),Min-Cheol Hong(홍민철) 대한전자공학회 2017 대한전자공학회 학술대회 Vol.2017 No.6
In this paper, we proposed a video stabilization algorithm to cluster the undesired motion features before smoothing by alpha-trimmed mean filter. Our method calculates the direction of movement from the features and applies directional statistics to find out the movement of moving objects and the movement of shaky to classify it. The process consists of three main steps: detecting the motion, clustering to get the good motion and smoothing the undesired motion by alpha-trimmed mean filter. The experimental results giving the effectiveness of our proposed method with low computation time to apply in real time system.
Motion net estimation and mixed norm filter for video stabilization
Minh-Thien Pham(팜민티엔),Min-Cheol Hong(민철홍) 대한전자공학회 2020 대한전자공학회 학술대회 Vol.2020 No.8
Video stabilization is to eliminate unwanted vibrations caused during video recording. In this paper, we introduce a motion estimation model using deep learning network instead of the traditional motion estimation method based on feature tracking to improve system performance as well as open other approaches to video stabilization using deep learning in the future. The motion network uses the Siamese network to receive input from a pair of adjacent RGB images from the video to estimate geometric affine transform that mapping the first image to the second image. Besides, the proposed model also detects the correlation between pairs of adjacent images to decide whether they exist or not. This is applicable in cases where the input videos do not have a continuous context