Experiments on improving the compression of special data types

Summary form only given. This paper, reports on experiments with two data compression schemes. The first seeks to improve the fidelity of reconstructed images through the parallel application of lossless and lossy compression to disjoint fields of image pixels. The second experiment is based on the observation that many relational database records exhibit a strong locality behavior, i.e., consecutive characters (within the same attribute) tend to be of the same type (alphabets, digit, trailing blanks, successive zeros, etc.). Adaptive compression schemes may not be quick enough in capturing and adapting to the transition from one field to the next. These can be modified by splitting the original alphabet into two or more smaller alphabets