A method of compressing data comprising separating a bit stream of compensation data into a sub-bit stream including an impulse component and a sub-bit stream not including the impulse component, predicting and calculating a trend component of compensation data to be compressed using compressed compensation data, generating a noise component by eliminating the trend component from the compensation data, and separating and compressing an impulse component and a noise component of the compensation data from each other, wherein the impulse component is compressed without quantization and the noise component is quantized and compressed.