MULTIBEST: multiple vector interpolation for vector-quantization based image compression

The paper reconsiders the applicability of Vector Quantization (VQ) to image compression for low bit-rate image transmission. The proposed method overcomes the basic, structural drawbacks of VQ by a general multiple-interpolation mechanism. The major advantages of the described schema are an improved generalization performance and a notable reduction in coarseness. The overall approach can then be integrated with classical adaptive methods to derive a flexible and effective compression schema without affecting compression performances. Massive experimental results on real and artificial images demonstrate the model's notable advantages over classical VQ systems and DCT-based standards.