Multiresolution Motion Estimation for Low-Rate Video Frame Interpolation
- PDF / 1,602,695 Bytes
- 13 Pages / 600 x 792 pts Page_size
- 39 Downloads / 234 Views
Multiresolution Motion Estimation for Low-Rate Video Frame Interpolation Hezerul Abdul Karim Faculty of Engineering, Multimedia University, 63100 Cyberjaya, Selangor, Malaysia Email: [email protected]
Michel Bister Division of Engineering, The University of Nottingham, Malaysia Campus, Wisma MISC, 50450 Kuala Lumpur, Malaysia Email: [email protected]
Mohammad Umar Siddiqi Faculty of Engineering, Multimedia University, 63100 Cyberjaya, Selangor, Malaysia Email: [email protected] Received 22 October 2003; Revised 18 February 2004; Recommended for Publication by C. C. Jay Kuo Interpolation of video frames with the purpose of increasing the frame rate requires the estimation of motion in the image so as to interpolate pixels along the path of the objects. In this paper, the specific challenges of low-rate video frame interpolation are illustrated by choosing one well-performing algorithm for high-frame-rate interpolation (Castango 1996) and applying it to low frame rates. The degradation of performance is illustrated by comparing the original algorithm, the algorithm adapted to low frame rate, and simple averaging. To overcome the particular challenges of low-frame-rate interpolation, two algorithms based on multiresolution motion estimation are developed and compared on objective and subjective basis and shown to provide an elegant solution to the specific challenges of low-frame-rate video interpolation. Keywords and phrases: low-rate video frame interpolation, multiresolution motion estimation.
1.
INTRODUCTION
Low frame rates, for example 10 frames per second (fps), are of interest in low-bit-rate video compression. Reducing the frame rate to 5 fps and interpolating it back to 10 fps at the receiver helps to reduce the transmission bit rate. In order to achieve very low data rates for video communication such as in plain old telephone service, it is necessary to skip images at the transmitter, which then have to be reconstructed at the receiver end. The reconstruction of the images can be achieved through frame interpolation. Interpolation of video frames simply means inserting or adding new frames between the video frames. Given the previous and next frames, the task is to insert a new frame between the two. In general, frame interpolation can be performed as illustrated in Figure 1. Without motion estimation, video frames can be interpolated by averaging the previous and next frames or by repeating the previous frame. The performance of frame interpolation will improve if motion estimation is included in the process. Motion estimation is used to estimate the motion
vectors between frames. Pixels are then interpolated along the path of the motion vectors. In recent years a number of frame interpolation algorithms have been developed [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16]. Most of them concentrate on high-frame-rate video as shown in Table 1 [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13] and part of [16]. In such cases, motion estimation can be achieved by simple block-matching techniq
Data Loading...
