Auroral Lossless Compression Device and Method

Lossless Compression of Auroral Spectral Images Using Inter-band Adaptive Linear Prediction

This technology is a device for the lossless compression of auroral spectral images consisting of M bands and K spectral lines. A comparator generates a prediction error based on the difference between the measured pixel value and the adjacent band predicted pixel value, while a quantizer, a 4th-order linear predictor, and an encoder convert this into bits for transmission over a channel.

Auroral spectrometers at Antarctic stations generate 1024×1024 16-bit images every 20 seconds, but multiple observatories have historically shared a single satellite channel with limited bandwidth. Lossy compression compromises high-precision scientific data, and general-purpose lossless methods like JPEG-LS or JPEG2000 have limitations in real-time transmission because they do not fully utilize spectral correlation.

This technology leverages both spatial and spectral correlations and calculates prediction coefficients through recursive matrix updates, eliminating the need to send additional information about the coefficients. It can be applied to data transmission systems for polar observatories, satellite remote sensing, and space weather observation equipment, enabling the real-time acquisition of original-quality observation images even over narrow communication bandwidths.

‍

‍

Key Features:

• A comparator that outputs a prediction error signal based on the difference between the measured pixel value of the m-th band and the predicted pixel value of the m+1 band for the k-th spectral line
• A quantizer that quantizes the prediction error signal, and an encoder that assigns bits to the quantized signal and transmits it to a channel
• A first predictor that generates the predicted pixel value for the m+1 band using a linear combination of 4th-order prediction coefficients and measured pixel values from adjacent spectral lines
• A reconstruction configuration comprising a decoder that decodes the encoded signal, an adder that restores the measured pixel value by adding the predicted pixel value, and a second predictor

‍