one idea is to do arithmetic coding of the angular displacements at each step along the whisker. The distribution of angular steps should be pretty tight, especially if you condition on the previous angle (a markov chain). This would give a pretty good encoding.
I have an arithmetic coding implementation, but it doesn't do markov chains yet. There's also some important details to be worked out about exactly what should be encoded; angle might not be the best thing.
one idea is to do arithmetic coding of the angular displacements at each step along the whisker. The distribution of angular steps should be pretty tight, especially if you condition on the previous angle (a markov chain). This would give a pretty good encoding.
I have an arithmetic coding implementation, but it doesn't do markov chains yet. There's also some important details to be worked out about exactly what should be encoded; angle might not be the best thing.