Laplacian Pyramid-based Downsampling Combined With Directionlet Interpolation For Low-bit Rate Compression

We present a novel image compression framework after the brief analysis to the theory and practical application about the current situation and future trends. Based on that we give the introduction of this new low-bit-rate image compression method from the perspective of interpolation reconstruction. On the encoder side of the framework, we can get the low resolution image by down-sampling the original, and then encode the low resolution one to obtain the compression bit stream. On the decoder side, we can get the reconstructed low resolution image by decoding the compression bit stream, and then interpolate the low resolution one to get the estimation of the original. The most important part of the whole framework is the interpolation module.This article focused on the Laplacian down-sampling, the directionlet interpolation, as well as the reconstruction of low-bit-rate image compression framework of such issues. The performance comparison between our proposed method and JPEG2000 standard method is presented not only in objective quality, but also subjective visual quality still gives out in this article under conditions of low-bit-rate.Innovative research work and related important results are as follows:1. The low-resolution image representation of Laplacian down-sampling is more in line with the real situation: In this article, Laplacian down-sampling is suitable for directionlet interpolation, so the combining of the Laplacian down-sampling and directionlet interpolation can achieve better reconstruction results. This reflects the supportive of the Laplacian down-sampling to the directionlet interpolation.2. Directionlet interpolation can be used for reconstructing the original effectively through the way of interpolating the low resolution image obtained by Lapalacian downsampling: In the directionlet interpolation, we applied POCS algorithm and the estimated Direction Map to the super-resolution image reconstruction. In this paper, we make reference to the implementation of the POCS algorithm, and learn the principle of the Initial Estimate & Iterative Amendment super-resolution image reconstruction thinking, and then give concrete steps to achieve directionlet interpolation. The final results show that directionlet interpolation can significantly improve the quality of reconstructed image.3. Low-bit-rate image compression framework is independent with the ADL and the TDLT coding methods, and is superior to the JPEG2000 method under the same bit-rate: On the encoder side, the low-resolution images obtained by the Laplacian down-sampling, is compressed by ADL(Adaptive Directional Lifting-Based Wavelet Transform) or TDLT(Time Domain Lapped Transform) to retain texture and details. On the decoder side, the low-resolution images decompressed by ADL or TDLT, is interpolated by the directionlet to reconstruct the original. The final results show that under the low-bit-rate conditions, this method is not only in objective PSNR values and subjective visual performance that superior to JPEG2000, but also the compression framework is independent to the selection of coding methods.