A Weighted Image Compressive Sensing Method Based on General Hidden Markov Tree Model

Abstract

The invention provides a weighted image compressed sensing method based on a universal hidden Markov tree (uHMT) model. According to the method, an iteration weighting L1 minimization algorithm based on the uHMT model is applied to compressed sensing reconstruction of an image, an HMT (Hidden Markov Tree) model is replaced by the uHMT model according to the characteristics of a natural image, a model parameter is given directly, and a training step with a larger computation amount is omitted, so that the computation complexity of the image reconstruction is reduced effectively. In addition, a computation method of a weight number influences the reconstruction precision greatly, and with the adoption of the improved weight number computation method by the weighted image compressed sensing method, the uHMT model describing a wavelet coefficient is introduced to the image reconstruction, and the incidence of a rejection model is reduced, so that the reconstruction precision of the image is improved.

Description

technical field [0001] The invention relates to the field of image compression, in particular to a weighted image compression sensing method based on a general hidden Markov tree model. Background technique [0002] With the advent of digitalization and the information age, the amount of image data is constantly expanding, which brings enormous pressure to image transmission and processing, so it is necessary to effectively compress and reconstruct images. In recent years, Donoho et al. proposed a new theory - Compressed Sensing Theory (CS). This theory breaks through the limitation of sampling rate in traditional Nyquist sampling. The sampling rate is no longer determined by the bandwidth of the signal, but by the structure and content of the signal. Its core idea is to combine the sampling and compression process in the traditional compression process, randomly project the signal to obtain a small number of observations, and then reconstruct the original signal from the o...

AI Technical Summary

Problems solved by technology

[0038] In order to solve the problem of high computational complexity of HMT model parameter estimation, combined with the characteristics of natural images, the present invention uses uHMT model instead of HMT. The uHMT model directly provides model parameters without the need to train model parameters through EM algorithm, that is, uHMT model The training steps with a large amount of calculation are omitted, and the calculation complexity is effectively reduced; at the same time, in view of the second shortcoming of the prior art, the improved weight calculation method provided by the present invention introduces the HMT model into the reconstruction algorithm and at the same time It also reduces the possibility of wavelet coefficients occasionally repelling the model, thereby improving the accuracy of the reconstruction value and improving the reconstruction quality of the image

Images