Multi-Spectral Compression Algorithm Based On Clustering And The Wavelet Transform

Multi-spectral images have a wide range of applications in environmental detection,biomedicine,and military investigation.It is a collection of spatial reflections of the same object under different spectral bands.It also has 1-D spectral information and 2-D spatial information.The amount of information makes multi-spectral images require a huge amount of resources for transportation and transmission.Therefore,the research of multi-spectral image compression algorithms is imperative.The development of multi-spectral image compression algorithms relies on spectral imaging techniques.This paper first introduces the origin of spectral imaging technology,and analyzes the history of spectral imaging at home and abroad.At the same time,it discusses the ideas of three commonly used multi-spectral image compression algorithms and the research content of domestic and foreign scholars;and it also discusses multispectral images.The evaluation system of compression quality was specifically discussed,laying the foundation for the subsequent multi-spectral image compression theory research.This thesis also analyzes the statistical characteristics of multi-spectral images,defines the spatial redundancy factor and the inter-spectral redundancy factor to represent the spatial redundancy of multi-spectral images,and quantitatively analyzes the coefficients of the two spectra.It is proved that multi-spectral images are available.Strong spatial redundancy and spectral redundancy.In order to remove the spatial redundancy and spectral redundancy,this thesis introduces some basic theory of wavelet and clustering,introduces the construction principle and development history of wavelet transform,constructs the wavelet basis and some commonly used image processing.The principle of wavelet basis was explained.At the same time,the theoretical basis of clustering was analyzed.The algorithm of the spectral clustering algorithm adopted in this paper was designed;and the JPEG2000 standard,which is often used in image compression,was taken as the core of wavelet transform.Instructions.Based on the above theoretical research,a multi-spectral image compression algorithm based on improved spectral clustering and wavelet transform is proposed in this thesis.That is,the spectral redundancy between multi-spectral images is removedby an improved spectral clustering algorithm,and the class-like representative images are classified.Wavelet transform is performed,and the output is coded by SPIHT algorithm.Karhunen-Loeve(KLT)transform is used to reduce the dimension of differential components between classes.Finally,wavelet transform and coding are performed to output the code stream.At the same time,in order to improve the accuracy,the image represented by the class-representing code stream is adaptively adjusted,and the quantization error generated in the encoding process is reduced.Experiments show that compared with the JPEG2000 and KLT+DWT algorithms,the peak signal-to-noise ratio of this method is increased by about 7.2 dB at 30 bands,with better peak signal-to-noise ratio and compression ratio,and it is suitable for compression of different spectrum segments.The strong stability meets the requirements of Mars multispectral camera compression and has broad prospects in practical applications.