Image Segmentation and Residual Coding
for Very Low Bit Rate Compression
Dorin Comaniciu and Peter Meer
Department of Electrical and Computer Engineering
Rutgers University, Piscataway, NJ 08855, USA
To achieve very low bit rate compression a new robust segmentation
method is employed to define the most important regions of the
image. The region contours are then approximated with polygonal lines
and transmitted together with the color of the region. The discrete
wavelet transform is used to compress the residual image which
represents mostly texture and noise. The overall procedure proves to
be an effective compression scheme. The reconstructed images show
satisfactory visual appearance at compression ratios
higher than 100:1 for gray level images and 150:1 for color images.
