Mahidol University Logo
Faculty of ICT, Mahidol University
 

Admissions

Printable Version

 

FRAME RATE UP CONVERSION BY BLOCK BASED AFFINE TRANSFORM

 

TITLE FRAME RATE UP CONVERSION BY BLOCK BASED AFFINE TRANSFORM
AUTHOR SARANTA PONLA
DEGREE MASTER OF SCIENCE PROGRAMME IN COMPUTER SCIENCE
FACULTY FACULTY OF SCIENCE
ADVISOR SUKANYA PHONGSUPHAP
CO-ADVISOR SUPATANA AUETHAVEKIAT
 
ABSTRACT
Frame rate up conversion is applied to video sequences to increase their frame rates and produce the smoother motion of objects. In order to produce the motion that is natural to human perception, an accurate description in terms of motion parameters is necessary. Because motion in video sequences consists of not only translation but also rotation and zooming, affine transform is more suitable to describe the motion than the conventional 2D translation model. This thesis proposes a frame rate up conversion based on affine parameters. An image in each frame is divided into blocks of 16x16 pixels. Frame difference is used to separate the background from foreground. Motion estimation based on modified three-step search is applied to find the affine parameters of the foreground blocks. Overlapped processing is applied. For each block of 16x16 pixels, the motion is used to describe the motion of the center 8x8 block. After the affine parameters are acquired, they are interpolated for the motion to the frame between two available frames. Because the changing rate of affine parameters is not linear, this thesis proposes the approximation technique for interpolating the affine parameters based on rigid transform. The frame in the middle of two available frames is reconstructed by moving foreground blocks according to the interpolated affine parameters. The background block is copied to the new frame. It is not guaranteed that every pixel can be reconstructed. Thus, a hole filling algorithm based on the motion parameters between the available frame and the reconstructed frame is applied. Experiments on various standard video sequences indicated that in video sequences without complex motion, there was no distinct difference between the reconstruction by 2D translation parameters and affine parameters. Reconstruction by 2D translation parameters produced slightly higher PSNR. However, in the video sequence with complex motion, the reconstruction based on affine parameters produced higher PSNR and provided smoother motion than the reconstruction based on 2D translation parameters.
KEYWORD FRAME RATE UP CONVERSION/BLOCK BASED MOTION ESTIMATION/ IMAGE INTERPOLATION

 

Go to Top

 

ICT Building, Mahidol University, 999 Phuttamonthon 4 Road, Salaya, Nakhonpathom 73170 Tel. +66 02 441-0909 Fax. +66 02 849-6099
Mahidol University Computing Center, The Faculty of ICT, Mahidol University , Rama 6 Road, Rajathevi, Bangkok 10400 Tel. +66 02 354-4333 Fax. +66 02 354-7333