Improving vector quantization of satellite images through the application of bi-orthogonal wavelets

In image coding applications, vector quantization (VQ) has widely been accepted as an effective compression method for digital images. However, its effectiveness depends greatly on the matching of the VQ elements with the image data values. Prior to the VQ process, orthogonal transforms such as a DCT or FFT are often used to convert an image in the spatial domain to transform coefficients in the frequency domain. In this paper, instead of orthogonal transform coefficients, bi-orthogonal wavelet filters are used to transform a satellite image to the scale-frequency domain. The wavelet coefficients are vector quantized and their statistical features are analyzed in details such as computational complexity and performance efficiency. Some examples will be shown in the presentation to illustrate the advantage of using bi-orthogonal wavelets that can help to improve the matching of VQ elements to the image data values. Through matching the image values, it can be shown that fewer VQ elements would be required to represent the image while maintaining the image quality, Improving the VQ statistical features will in turn increase the compression ratios of image compression of satellite images.