| Abstract |
Remote sensing centers have currently more and more available satellite images of the earth surface. The satellite images produce large volumes of information, and each image may have a size that ranges from tens of megabytes to a gigabyte. So, compression can help to reduce the huge storage requirements needed. On the other hand, global positioning systems (GPS) provide the locations and information for users. Therefore, using satellites for navigation has become an indispensable part of modern life and the transmission of these satellite images have become one of the greatest challenges. Moreover, hyperspectral images have been used in a wide range of applications and remote sensing projects ranging from independent land mapping services to government and military activities. Some bands of the hyperspectral satellite image are corrupted by noise due to the atmosphere and the channel noise between the satellite and the ground station. Therefore, when we have more than one version of the same hyperspectral image, we can combine theme to achieve the best image quality. Thus, the need for development of new communication schemes for hyperspectral satellite image becomes more and more imperative to improve the quality of the received images.
This work is focused on the efficient communication schemes of the satellite imagery. In the communication system, there are two different parts; compression and transmission. The first part of this thesis introduces two lossy compression techniques using wavelet transform and rate control for satellite images. The first compression technique is based on removing some wavelet sub-bands from the most correlated bands. The removed sub-bands are determined using the correlation coefficients between bands and then they are reconstructed at the decoder from the most correlated sub-bands with them. This method reduces the size of storage of satellite images while keeping high-quality reconstruction. Enhanced Thematic Mapper plus (ETM+) satellite multispectral images are used for evaluation. Simulation results demonstrate that the proposed method improves the average multispectral image quality by 0.5dB to 9dB. The second compression technique introduces a simple trend for lossy compression of satellite images based on rate control aiming at decreasing the impact on vegetation features. This compression technique is based on a new rate control scheme in which the vegetation group (i.e., group containing the most significant bands for vegetation feature extraction) is encoded at a higher bit rate than that of the remaining group. The study is performed on several satellite images, where multispectral images are selected from an ETM+ satellite, and hyperspectral images are selected from an Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) satellite. In all simulation results, the traditional compression methods exhibit more effects on the vegetation features of satellite images. Besides, the proposed method has a less impact on vegetation features. These results validate the effectiveness of the proposed method.
The second part of this thesis introduces two wireless communications schemes for broadcasting high resolution satellite images to a large number of receivers based on SoftCast and distributed coding. These schemes are extended to deal with the hyperspectral satellite images. The first communication scheme, called HyperCast, uses a clustering transformation approach. This approach divides the hyperspectral image bands into groups, and then employs the transforms to remove the redundant information from each group. After that, the transmission power is directly allocated to the transformed data according to their distributions and magnitudes without forward error correction (FEC). These data are transformed by Hadamard matrix and transmitted over a dense constellation. HyperCast achieves in broadcasting a PSNR gains up to 6.98dB, 3.48dB and 6.14dB over LineCast, SoftCast-3D, and the conventional framework, respectively. The second communication scheme is based on wavelet transform and distributed source coding (DSC). Before wavelet transform, the pre-processing is performed in two steps which are hyperspectral band ordering and normalization. In the wavelet transform, the DWT with three-level decomposition is used to divide each hyperspectral image band into ten wavelet sub-bands; nine of them are the details and the last LL-LL-LL is an approximation version of the band. Coset coding based on DSC is used for the LL-LL-LL sub-band to achieve high compression efficiency and low encoding complexity. The detail wavelet sub-bands and the coset values are transmitted as in the previous proposed scheme without FEC. Experimental results on broadcasting demonstrate that the proposed scheme improves the average image quality by 6.91dB, 3.00dB and 5.63dB over LineCast, SoftCast-3D, and the conventional framework, respectively. In addition, the proposed scheme has less encoding time than all state-of-the-art schemes. Finally, the proposed communication schemes present efficient hyperspectral band ordering algorithms based on the correlation coefficients and absolute differences between bands to improve the compression performance. The proposed band ordering algorithms achieve better performance as compared to the case of no band ordering up to 1.30dB on average PSNR.
|
