Optimal quantization and bit allocation for compressing large discriminative feature space transforms | IEEE Conference Publication | IEEE Xplore