论文部分内容阅读
本文用矢量量化技术将成象光谱仪的三维谱象数据空间景象中每一象元对应的光谱定义为一个矢量，用基于光谱特征的二进制光谱编码方法对各光谱矢量进行编码，用编码后的光谱矢量来进行快速码字匹配．这种三维谱象数据压缩方法不仅使处理速度大大加快（当码书取２５６码字时比常规的矢量量化方法快３０倍，码书取４０９６码字时快４３倍），而且压缩后恢复图象精度还有所提高．本文定义的矢量构成方法不仅可有效地保存光谱特性，而且还可充分利用成象光谱数据光谱维的相关性，可获得相当诱人的压缩比．当压缩比高达１９２：１时恢复数据仍可达到４５．２ｄＢ的峰值信噪比．
In this paper, the vector corresponding to each pixel in the image space of the imaging spectrometer is defined as a vector by vector quantization, and each spectral vector is encoded by a spectral spectral feature based binary spectral encoding method. The encoded spectral vector For fast codeword matching. This three-dimensional spectral data compression method not only speeds up the processing greatly (when the codebook takes 256 codewords, it is 30 times faster than the conventional vector quantization method and 43 times faster when the codebook takes 4096 codewords), and the compressed recovery graph As the accuracy has improved. The vector construction method defined in this paper can not only preserve the spectral characteristics effectively, but also make full use of the correlation between the spectral dimensions of the imaging spectral data to obtain a rather attractive compression ratio. When the compression ratio up to 192: 1 data recovery can still reach 45.2dB peak signal to noise ratio.