Distortionless data compression of NMR images using vector quantization

An application of vector quantization (VQ) to subjective distortion-free compression of two-dimensional NMR (nuclear magnetic resonance) image data is presented. An alteration of regular VQ is used to allow a linearly growing codebook size which adaptively compresses the input data, resulting in a reduced-size output image file. Due to certain subjective constraints of the biomedical profession, distortionless data compression is sometimes required to store, transmit and reconstruct the original signal. With the aforementioned VQ method, the compressed signal is composed to components: the VQ codebook index and the actual error vectors between the input and the nearest-match codeword. Distortionless compression ratios of 10:1 were realized for test data. The algorithm worked sufficiently well for cranial, spinal skeleton, and organ tissue images.<
>