A Concealment Method for Shape Information in MPEG-4 Coded Video Sequences
Shahram Shirani, Berna Erol, and Faouzi KossentiniIEEE TRANSACTIONS ON MULTIMEDIA, VOL. 2, NO. 3, SEPTEMPER 2000
Outline
introduction MPEG-4: video compression and error
resilience concealment of channel errors in
MPEG-4 coded video proposed method experimental results
Introduction
MPEG-4 : Video Compression and Error Resilience
MPEG-4 offers error resilience tools:– video packetization– data partition– reversible VLC
MPEG-4 : Error Resilience
video packetization– The macroblock data between two resynchronization
markers is referred to as a video packet. data partition I-VOPs P-VOPs
reversible VLC
Concealment of Channel Errors in MPEG-4 Coded Video Data
P-VOPs
I-VOPs
Proposed method
Proposed method
The boundary transform:– if all the four neighboring pixels of a pixel are inside the
shape, then the pixel is declared inside. Otherwise, the pixel is assigned to the border of the shape.
Experimental Results
The similarity of the erroneous and error concealed shape data to the original shape data is measures using the following ratio:
– nd : the number of pixels that are different between the restored shape and the original one
– nt : the total number of pixels in the bounding box
Experimental Results (AKIYO)
272x208 30% missing
88%
99% (Mp-Mm)/M
= 17%
Experimental Results (BREAM)
144x192 29% missing
81%
(Mp-Mm)/M
98% = 22%
Experimental Results (BREAM)
272x192
25% missing
86%
99%
Experimental Results (WEATHER)
160x224
35% missing
87%
99%
Conclusions
The paper presents an error concealment method for shape information in the MPEG-4 coded video sequences.
The proposed concealment method successfully reconstructs the missing shape data providing about 20% improvement compared to median filtering method.
Reasonable computational complexity of our algorithm makes it suitable for real-time application.