Fractional wavelet filter based low memory coding for hyperspectral image sensors

In the present study, a novel low memory coding algorithm for lossless image compression of hyperspectral images is proposed. The hyperspectral images are volumetric images that pose a challenge to the sensor memory. The contemporary transform-based compression algorithms exhibit remarkably efficient performance on the coding gain, complexity, and memory in comparison to other algorithms for lossy compression. The traditional 3D-DWT requires large memory for computation of wavelet coefficients of transform image. The fractional wavelet filter is a low memory solution to calculate the wavelet coefficients of the hyperspectral image. The 2D-ZM-SPECK is employed as a coding algorithm which is applied over HS image frame by frame basis. The simulation results indicate that the proposed compression algorithm has low memory requirements and high coding gain with less computational complexity. On observing the simulation results of the proposed compression algorithm, it is noticeable that the proposed coder is fast enough due to requiring low memory and hence proving its candidature in the implementation of a resource-constrained hyperspectral image sensor.