Predictor XOR

The PredictorXor transform uses a fixed-context model (FCM) to predict elements within a plane, emitting an XOR residual that entropy coders can compress efficiently.

Theory

For each element in a plane, PredictorXor predicts its value from previous values using a small fixed-context hash table. It then emits truth XOR prediction.

When the underlying values change smoothly, the output residuals cluster around zero, exhibiting a heavy leading-zero distribution. Entropy coders like Huffman or FPC excel at compressing these distributions.

The predictor maintains a 64K-entry hash table indexed by a rolling hash of recent values. The decoder runs the exact same recurrence in lock-step to recover the original sequence bit-exactly.

Usage

The transform operates on Byte, Word2, Word4, and Word8 element widths. It requires no parameters, as the hash table size and shift parameters are fixed.

References

• Martin Burtscher and Paruj Ratanaworabhan, "FPC: A High-Speed Compressor for Double-Precision Floating-Point Data", IEEE Transactions on Computers (2009). Link