Browse Prior Art Database

Method for Reducing Pulsation in MPEG Video Coding

IP.com Disclosure Number: IPCOM000014940D
Original Publication Date: 2002-Jan-29
Included in the Prior Art Database: 2003-Jun-20
Document File: 3 page(s) / 84K

Publishing Venue

IBM

Abstract

Method For Reducing Pulsation In MPEG Video Coding Larry Lu, Jack Kouloheris, Cesar Gonzales What is disclosed is a rate control scheme that can be used to prevent the pulsation artifact in MPEG still scene video coding. Pulsation in MPEG video coding of still scenes is an annoying artifacts. In a still scene, each picture has the same signal characteristics. (There may be some noise, however it is often negligible in most cases.) Ideally, such still scene should be encoded consistently. However, if any portion of the picture is not coded consistently throughout the sequence, it may produce pulsing artifacts. Furthermore, MPEG video encoders typically code the incoming pictures with different coding types according to a pre-selected pattern, for example, the size of group of pictures (GOP) N=15 and the reference picture distance M=3. Therefore, when the video source changes from motion scene to still scene, whether there will be pulsation artifacts will largely depend on the encoder's rate control scheme. Since the pulsing artifact means that the picture qualities over the scene are not consistent, it implies that the bit allocation algorithm is not functioning appropriately. Therefore to eliminate the pulsation artifacts, we must consistently distribute the bits over same type of pictures and manage the VBV buffer at appropriate level to prevent any inconsistent bit allocation.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 3

Method for Reducing Pulsation in MPEG Video Coding

Method For Reducing Pulsation In MPEG Video Coding

Larry Lu, Jack Kouloheris, Cesar Gonzales

What is disclosed is a rate control scheme that can be used to prevent the pulsation artifact in MPEG still scene video coding. Pulsation in MPEG video coding of still scenes is an annoying artifacts. In a still scene, each picture has the same signal characteristics. (There may be some noise, however it is often negligible in most cases.) Ideally, such still scene should be encoded consistently. However, if any portion of the picture is not coded consistently throughout the sequence, it may produce pulsing artifacts. Furthermore, MPEG video encoders typically code the incoming pictures with different coding types according to a pre-selected pattern, for example, the size of group of pictures (GOP) N=15 and the reference picture distance M=3. Therefore, when the video source changes from motion scene to still scene, whether there will be pulsation artifacts will largely depend on the encoder's rate control scheme.

Since the pulsing artifact means that the picture qualities over the scene are not consistent, it implies that the bit allocation algorithm is not functioning appropriately. Therefore to eliminate the pulsation artifacts, we must consistently distribute the bits over same type of pictures and manage the VBV buffer at appropriate level to prevent any inconsistent bit allocation.

Obviously, still scene coding is very different from motion scene coding. The coding of still scenes requires that the pictures are encoded as consistently as possible, while the coding of motion scenes demands quick adaptation to the time varying characteristics of the non-stationary video signal to keep the picture quality as smooth as possible. Therefore, a good rate control scheme for motion scene coding may become inappropriate or even bad for still scene coding and cause severe pulsation artifact. Since our objective is to eliminate the pulsation artifacts in still scene coding while maintain good performance in motion scene coding, our first step is to find a way to detect on the fly whether the input pictures are still scenes so that an appropriate rate control scheme can be applied accordingly.

We use the pixel difference (PD) of the pictures to measure the variation degree of the pixel values from picture to picture. As we can expect, for still scene, PD should be small in value. So we compute the PD's between the current picture and previous pictures and compare the PD values with a threshold. If the PD's are greater than the threshold, the input sequence is a motion scene; otherwise it is a still scene. Once the still scene is detected, we need to apply an appropriate rate allocation scheme to prevent the pulsation artifacts. In still scene coding, first, the target rate mo...