The Transform Based Low Complexity And Low Memory Image Compression Method

A low computation and low memory needed image compression method is proposed in this thesis, which is based on the lapped biothorgonal transform (LBT) and simplified zero tree codec.As we know, discrete wavelet transform (DWT) based image compression method is prevalence now, which gets lower bit rate and better image quality. First, it is because of the advantages of wavelet transform, such as multiresolution and the ability to keep track of time/space and frequency simultaneously and represent local features better. At the same time, the proper codec fitting the special distribution of wavelet coefficent improves the coding efficiency, and generates the scalable bit stream, which can be used in various environments.But it must be pointed out that most wavelet based compression methods are of large amount of computation and cost space several times of that the image needs, which is possible in some particular conditions, such as in satellites and portable devices. And the wavelet transform itself is of so long basic functions that serious ghost appears in the decompressed images at low bit rate.To overcome the shortcomings of the wavelet based method, the lapped biorthogonal transform based low complexity and low memory image compression method is proposed in this paper, which needs much less computation and buffer. LBT costs 38% computation as the discrete wavelet based , and only buffers 1/15 data that DWT does. By analyzing the distribution of the coefficients of LBT, the quantized set partition coding method is proposed and used in the method. In coding process no context information is needed, and no complicated data structure, such as list, is used, so it is low memory and low computation. The compression results on many standard test images show that, using the same coding method, LBT based one performs better than DWT based, with higher PSNR and less distortion. We also find that LBT based method is especially fit to images with more details, such as satellite images. After all, our method is suitable to portable devices and real time processing.