Image Compression - Bezier and Splines in Image Processing and Machine Vision - page 82

perceived. However, to perceive difference between the two images, one can
use completely different values as we have done for Figure 4.1(c). Here, we
have chosen zero gray values (completely dark) for textured regions. To get
segmented images with texture regions, we first found the textured regions,
and then the non-textured segmented regions from the rest of the regions.
Their gray values have been replaced by respective threshold values. For an

(a) (b)

(c)

Fig. 4.1. (a) Input Lena image; (b) segmentation without texture region; (c) seg-
mentation with texture region.

input image of k -1 thresholds, we have k subimages. If N 1 ,N 2 , N k are the

number of sparse image surface patches in them, then considering variable
order global approximation of k subimages and variable order approximation
for sparse residual surface patches, the compression ratio for an M

···

× M image

with L graylevels becomes