Low-end 8-bit Compression: Mtf Too Poor, Adapthuff Too Expensive: Alternatives?

Hi! I am working on data compression, and one of my targets is 8-bit computers. Right now, on some versions, I'm doing extremely exceptional, but the main 16-bit version is doing horrible on one large text file. That's not the topic of this thread. I have some side compression techniques for purposes such as open source and very small 8-bit computers. One such uses MTF, while another uses Adaptive Huffman codes. The one that uses MTF requires very little memory but is doing pretty poorly: only .2% better overall than Exomizer, while the other produces a very good compression ratio, but, according to my calculations, the Huffman tables for decompression alone costs a little more than 6.5k. I'm looking for a moderate technique that is not too difficult to implement. Any ideas?

 

Maybe this will help. I am not willing to state all of my secrets, but I'm using the following techniques among others to compress files:

* I use Last16, where, if an LZ77 block is the same as a previous block, it is shortened to the times ago where it occurred.
* I use a modified version of Elias encoding that produces better results than the original.
* I have a technique that, if a particular literal was repeated recently, it is shortened to a reference to the last copy. I call this Placement Offset Basic. An extended version has ways to decrease the size of an occurrence while increasing the magnitude of the occurrence, but I'm not willing to say how. I'm not using this, as MTF does the job.

You may use these ideas in your code provided you mention them in the foreword of the programs' docs, state that they were from me and tell me via e-mail you're doing so and provide a link to the program online. If you have any other ideas to better my compression technique, I'm all ears.