基于变换编码和分布式编码的谱带排序优化的计算机卫星图像通信方法
Chapter 1 Introduction1.1 MotivationMultispectral images are generated by collecting several bands representingthe same area of the earth surface in different spectral sampling intervals, andhyperspectral images are generated by collecting hundreds of bands, as shown inFig. 1-2, having two dimensions that represent spatial position and one thatrepresents wavelength. However, at the time we gain high resolution spectralinformation, we generate massively large image data sets. Therefore, storage andtransmission of these amounts of data have bee one of the greatest challenges,due to the mon limitations of storage and transmission in the remote sensingscenario.An efficient coding of the original image can reduce the amount ofinformation needed to store the image. Besides, depending on the scenario, imagecoding techniques can be divided into lossless and lossy techniques. In the losslesstechniques, the image reconstructed after the decoding process is perfectlyidentical to the original image. However, lossless pression typically achiev es alimited pression ratio. In lossy pression, we can achieve a highpression ratio, which the lossy pression used a small number ofinformation to represent the original multispectral or hyperspectral satellite image.However, there is some distortion between the reconstructed and original images,where we can achieve a high pression ratio. Lossy pression coding mayaffect the features of the original satellite images, and hence there is a need toensure that the quality of the reconstructed images is still adequate for the intendedscientific use. Therefore, in this thesis, efficient satellite image pressiontechniques are presented aiming to reduce the size of storage of satellite imageswhile maintaining high-quality reconstruction with low impact on image features.........1.2 Fundamental ConceptsA satellite image is a three-dimensional matrix of sample values x(M, N, λ)with band size M, N in the spatial dimensions and λ in the spectral dimension. Thesamples have a valid range of values, which is referred to as bit-depth, setting thenumber of bits needed to represent each sample value. The main contributions of this thesis are an efficient technique for satellitemultispectral image pression aiming at reducing the size of storage ofmultispectral, while maintaining high-quality reconstruction, a novel technique forlossy pression of satellite images aiming at decreasing the impact onvegetation features, and two schemes for satellite hyperspectral imagebroadcasting over wireless channels using a linear transform between thetransmitted image and the original pixels luminance..........Chapter 2 Related Wors for the Compression and theTransmission of Satellite ImageryThis chapter briefly reviews the popular satellite image pressiontechniques and the selected satellite image wireless munication techniques.2.1 Recent Satellite Image Compression TechniquesRemote sensing data have bee enormously important for a myriad ofapplications addressed for the Earth observation. Recent sensors can cover largegeographical areas, producing images of unprecedented spectral and spatialresolution. For instance, the hyperspectral satellite can capture hundreds of spectralchannels at an acquisition bit depth of 16 bpppb close to 20 GB, daily. Hence, theneed for efficient pression techniques for remote-sensing data bees moreand more imperative to improve the capabilities of storage and transmission.Satellite image pression techniques can be divided into lossless and lossytechniques. In the lossless techniques, the image reconstructed after the decodingprocess is perfectly identical to the original image. However,lossless pression typically achieves a limited pression ratio. In[24], anoverview of several standards for remote sensing data pression is introduced.For lossy pression, there is some distortion in the reconstructed image paredto the original image, but a high pression ratio can be achieved[25, 26, 16, 13, 27, 14,28, 29, 30]. In the following two sections, we are concerned with the up-to-date satelliteimage pression and transmission techniques for our investigation..........2.2 Recent Wireless Communication TechniquesWireless image and video munication has been studied for a long time.Shannon concluded that two main issues are required to transmit data over wirelesschannels; source coding (data pression) and channel coding (FEC andmodulation scheme)[52, 53]. This conventional scheme with separate source codingand channel coding is based on Shannon‘s source-channel separation principle, andit is the most classic digital scheme. Source coding is designed independently ofchannel coding. Using this classical scheme, the source data can be transmittedwithout any loss of information, if the channel is point-to-point (i.e., unicastmunication). The channel quality is nown or can be easily measured at thesource by the selection for the optimal transmission rate for the channel and thecorresponding FEC and modulation[53]. However, for digital broadcast/multicast, acliff effect challenge is found, where each receiver observes a different channelquality. Thus, the bitrate selected by a conventional wireless image delivery schemecannot fit all receivers at the same time. If the image is transmitted at a high bitrate,it can be decoded only by those receivers with better quality channels, but it is notreasonable for receivers with worse quality channels. On the contrary, if it istransmitted at a low bitrate supported by all receivers, it reduces the performance ofthe receivers with better quality channels, and it is not optimal for performance. Inorder to overe the cliff effect, many researchers proposed different jointsource-channel coding (SCC) framewors for distributed image/video transmission[54, 55]. Except for these SCC wors, the transmission of distributed coded video isstill similar to that of the conventional scheme. In contrast to the separate design,there are many joint image/video coding and transmission schemes that have been proposed for wireless image/video multicasting. SoftCast[56, 57]is one of the analog approaches that are designed within the SCC framewor.........Chapter 3 Satellite Image Lossy Compression Techniques Using WaveletTransform............303.1 Introduction ............ 303.2 Satellite Lossy Image Compression Based on Removing Sub -bands ....... 313.2.1 Correlation Coefficient between Bands ............ 333.2.2 Selection of Correlated Bands ...... 343.2.3 Removal of Image Sub-bands....... 343.2.4 Band Ordering........... 353.2.5 PEG2000 Encoding with PCA .... 363.2.6 The Decoding Process ........ 363.2.7 Evaluation Results .... 373.3 Lossy Compression of Satellite Images with Low Impact on VegetationFeatures .... 443.4 Summary ....... 59Chapter 4 High Resolution Satellite Image Broadcasting Schemes .....634.1 Introduction ............ 634.2 Satellite Image Broadcast Based on Wireless Softcast Scheme ..... 644.3 Distributed Coding and Transmission Scheme for Wireless Communication ........ 734.4 Summary ....... 81Chapter 5 Hyperspectral Image Broadcasting Using a ClusteringTransformation...........825.1 Introduction ............ 825.2 HyperCast ..... 845.3 Evaluation and Results ...... 96Chapter 6 Hyperspectral Image Communication SchemeBased on Wavelet Transform Domain Distributed CodingThis chapter presents a novel munication scheme for hyperspectral imagebroadcasting over wireless channels. The presented scheme is based on wavelettransform domain distributed coding. This scheme avoids the annoying cliff effectcaused by the mismatch between transmission rate and channel condition. Moreover,distributed coding achieves high pression efficiency and low encodingplexity, explained in the following sections.6.1 IntroductionAs we mentioned in Chapter 5, the need of development new schemes forhyperspectral satellite image broadcasting bees more and more imperative toimprove the quality of the received image. Recently, there are many SCC schemescan used in the hyperspectral satellite image broadcasting such as SoftCast[57]andLineCast[59]. While, the SoftCast putational plexity is high due to the 3-Dtransformation and the LineCast does not exploit the spectral correlation betweenbands. As a result, a new distributed coding and transmission scheme is proposed inthis paper for broadcasting satellite hyperspectral images to a large number ofreceivers. The scheme avoids the cliff effect found in the digital broadcastingschemes by using linear transform between the transmitted image signal and theoriginal pixels luminance. In the proposed scheme, the transform coefficients aredirectly transmitted through a dense constellation after allocating a certain powerwithout FEC and digital modulation. In multicasting, each user can optimal qualitymatching for its channel conditions. A DSC is applied based on the Slepian-Wolftheory[31]to achieve efficient pression and low encoding plexity. Moreover,a simple and more efficient algorithm is proposed for satellite hyperspectral band ordering to achieve better pression efficiency. The original satellite image bandordering is used in SoftCast, and thus, it is inefficient in the removal of spectralredundancy. Furthermore, the 3-D DCT which has been used is insufficient forremoving most of the redundancy in the satellite images. In contrast to SoftCast, theproposed scheme adopts a new band ordering algorithm to remove spectralredundancy in the hyperspectral image. Moreover, the DSC provides lower encodingplexity than those of traditional pression techniques. LineCast does notexploit the spectral correlation between bands.
..........ConclusionsIn this thesis, several contributions have been presented that advance the fieldsof satellite image munication. The contributions are focused on reducing thesize of storage of satellite images while maintaining high-quality reconstruction anddecreasing the impact on image features, and developing novel municationschemes for satellite image broadcasting over wireless channels. The majorcontributions of this thesis are summarized below:(1) An efficient multispectral image pression method based on removingsub-bands has been proposed in the first part of chapter 3. The proposedtechnique uses the correlation coefficient to determine the three mostcorrelated bands. After that, the DWT is used to determine the removedsub-bands from them. Finally, the remaining bands are pressed usingPEG2000 with PCA as a spectral decorrelator. We tae full advantage ofsize reduction by removing some wavelet sub-bands. Experiments resultsdemonstrate that the proposed method improves the average multispectralimage quality by 0.5dB to 9dB.(2) A novel pression method for satellite images with low impact onvegetation features has been proposed in the second part of chapter 3. It isbased on a new rate control strategy. The proposed method divided thespectral bands into two groups (i.e., vegetation and remainder groups) andencoding each of them at different bit rates. We also considered the bandordering algorithm for both groups in the pression process. The NDVIand the NDWI have been used to study the impact of lossy pressiontechniques on the satellite image features. The experimental results onmultispectral and hyperspectral images validate the effectiveness of theproposed method...........References (abbreviated)