Summary: | 碩士 === 國立臺灣科技大學 === 電子工程學系 === 84 === For applications such as video conference or videophone , it is
commonly agreed that foreground information is more important
then background . In conventional video coding schemes , shch
as H.261 , due to the variations of background lighting or
noise , a large amount of bits are used to code the background
areas . Such inefficient coding result in poor subjective
visual quality in the foreground areas of the reconstructed
image. In this thesis, we propose a video compression algorithm
for very low bit rate coding based on wavelet transform and
human visual characteristics . First , discrimination of
foreground and background areas is carried out on the residual
image obtained form the current frame and previously
reconstructed frame in the wavelet transform domain . Motion
estimation and compensation is made in the foreground areas of
the three lower subband . It is shown by expriments that the
proposed coding algorithm is capable of compressing video
sequences with reduced block effect and bet- ter visual quality
in very low bit rate applications .
|