Authors : Haitham S. Hasan and Mais A. Alsharqi
Abstract: Several proposed methods related to Hyper-Spectral (HS) image compression have been published in the recent years. These methods have often effective compression accuracy but they are time-consuming. This study introduces the development of a real-time practical scheme for use in lossy HS image compression. This scheme includes two parts; hardware using the Field Programmable Gate Array (FPGA) system and software utilizing the band prediction and fractal encoding techniques. The software technique starts by partitioning the HS image into a number of Groups of Bands (GoBs). Then, the first band in each GoB is utilized by the intra-band prediction to exploit the spatial correlation. And the other bands in each GoB are employed by the inter-band fractal coding technique as well as a limited search algorithm to make a complete benefit from the local matching between any two neighboring bands. This technique shows that the reconstructed image has a better improvement in the classification accuracy than the primary uncompressed image but still time-consuming. So, to overcome this problem, the technique is implemented by using the FPGA. This hardware technique is extremely suitable for real-time purposes.
Haitham S. Hasan and Mais A. Alsharqi, 2019. Lossy Compression of Hyperspectral Images Using Real-Time Technique. Journal of Engineering and Applied Sciences, 14: 4430-4434.