as a part of my diploma thesis I want to implement an LZ-based compression scheme. I need it to be statistical, i.e. it does not itself compress, but returns a probability for the next symbol. Also the implementation should work bitwise. (Since I am going to compress text, the compression scheme could internally work in a bytewise manner, but I need a prediction for every new bit.) 100 MB of memory should be sufficient.
There is a whole lot of different implementations both for LZ77 and LZ78. I dont know which is better. Could anyone suggest me a scheme, that is appropriate for my task? Maybe you can tell me the title of a paper for this scheme.
Thanks in advance!