Research On WMSN Video Denoising Based On Low-Rank Decomposition And MCA

Wireless Multimedia Sensor Networks(WMSN) is an important technology to active perceiving the physical world in the Internet of Things. WMSN technology has been widely used in Intelligent Transportation System and Industrial control in terms of the unique advantage of perceiving multimedia information. However, the complexity and high noise level of the monitoring environment lead to the result that WMSN videos often contain Gaussian noise and sparse noise such as rain streaks, impulse noise, etc, which seriously affect the video quality. Therefore, removing the mixed noise in WMSN video becomes the key to ensure the validity and reliability of the WMSN video monitoring.Recently, video denoising method based on low-rank matrix decomposition has drawn an increasing number of research attention. The main idea of low-rank matrix decomposition is decomposing data matrix into the low-rank matrix and sparse matrix, and to get the low-rank matrix by solving the nuclear norm optimization problem. Low-rank matrix decomposition theory can effectively remove sparse noise in WMSN video. In addition, Morphological Component Analysis(MCA) algorithm has a greater ability to remove many noises, including sparse error, streaks, etc. Morphological Component Analysis algorithm is proposed on the basis of sparse representation theory, which employs morphological differences of different elements in signal to separate different elements. Combining with the WMSN video and making thorough analysis of low-rank matrix decomposition and morphological component analysis algorithm, it will be of great use to study suitable denoising algorithm to remove mixed noise for WMSN video under complex scenarios.In practice, the mixed Gaussian-rain noise is one of the most common noises, which impairs the visibility or interpretability of the video. In this paper, the mixed Gaussian-rain noise removal method based on MCA-RPCA in WMSN is proposed. Firstly, WMSN video is decomposed into three parts: low-rank part, Gaussian noise part, and sparse part via Inexact RPCA. Secondly, sparse part is decomposed into rain component and nonrain component by performing dictionary learning and sparse coding based on MCA algorithm. Lastly, the noise-removed version of WMSN video can be gained by integrating the nonrain component of sparse part with the low-rank part of WMSN video. Experimental results show that the performance of the proposed approach is competitive, qualitative, and has greater ability to retain video feature information.Video data, as the main form of WMSN information, will consume large amounts of energy during transmission, which brings challenge to WMSN with limited energy of the node. In order to decrease computing time, reduce network time-consuming, we proposed the mixed Gaussian-rain noise removal method based on Improved MCA-RPCA in WMSN. A strong spatial-temporal correlation and redundancy among the adjacent frames are utilized in the proposed method. Firstly, WMSN video is decomposed into three parts: low-rank part, Gaussian noise part, and sparse part via Inexact RPCA. Secondly, we extract a set of overlapping patches from the key frame in the HF parts of the sparse part for learning dictionary, and partition the dictionary. Lastly, sparse part is decomposed into rain component and nonrain component by Improved Morphological Component Analysis(IMCA) algorithm, and the noise-removed version of WMSN video can be gained by integrating the nonrain component of sparse part with the low-rank part of WMSN video. Experimental results showed that the method can achieve better visual quality while removing much noise and further reduce the computational complexity, suitable for the WMSN video denoising.