Algorithm Of Integer Time-Domain Lapped Transform And Its Application In Video And Image Compression

The theory of time-domain lapped transform and its application in image and video compression are considered in this paper.Video compression at low bit-rates has become the focus of research in the signal processing community due to the expanding applications in video conferencing on N-ISDN and wireless video communications. Unfortunately, in these situations, the notorious blocking artifacts resulting from block DCT set a severe limit on the achievable bit-rate with acceptable quality. Recently, discrete wavelet transform (DWT) based compression method is prevalence , which gets lower bit rate and better image quality. But most wavelet based compression methods are of large amount of computation and cost space several times of other traditional methods, which is impossible in some poor conditions.The implementation of integer biorthogonal time-domain lapped transform (binTDLT) using lifting scheme and binDCT is proposed which is in the forms of shifts and adds and costs only 41% computation as the binDWT. A novel video compression algorithm based on the binTDLT and overlapping block motion estimation/compensation is presented. The proposed algorithm combining with simplified zero tree codec is of low computation and memory requirement while retaining the flexibility that block-based methods possess. Experimental results show that blocking artifacts has been effectively mitigated at low bit-rates in comparing with DCT based method and the proposed coding scheme is better than 3D-SPIHT for the scene of video sequences that move rapidly.A lapped filtering method is introduced to reduce the reconstruction errors near block or group boundaries in image and video compression. A low complexity remote sensing image compression method via lapped filtering based on spaced domain resampling named LFRC is proposed, which is characterized with high fidelity of the restored image and low computation complexity of the compression algorithm. Image coding experiments show that the LFRC performs better than the RBC, with higher PSNR and less distortion,moreover,it is superior to the SPIHT on preserving the edges and details of the image. Therefore,the LFRC is quite suitable for high speed and real time remote sensing image compression.