Conclusions
Simple method can be used to identify select and replace document regions without decompressing it
Segmentation parameters can be set based on a set of assumptions
Time for all processing is comparable to time spent only for full decompression.
Segmentation algorithm itself is an very fast illustrative example which can be replaced by any more sophisticated technique