Wireless Multimedia Sensor Network Image De-Noising via a Detail-Preserving Sparse Model

Abstract

In this paper, we propose a Detail-Preserving Sparse Model (DPSM) for de-noising of images that are usually interfered by noise on the Wireless Multimedia Sensor Network (WMSN). Specifically, based on the Structural SIMilarity (SSIM), the DPSM first incorporates a structural-preserving constraint, which enables the structure in the reconstructed image to be close to the ideal no-noise image. In addition, the DPSM adopts a residual ratio as the stopping condition of the sparse solution algorithm (e.g., Orthogonal Matching Pursuit), which enables the structures to be reconstructed under high noise conditions. The experimental results on several WMSN images have demonstrated the superiority of the proposed DPSM method over several well-known de-noising approaches in terms of PSNR and SSIM.

Keywords

• structural similarity
• residual ratio
• wireless multimedia sensor network
• sparse representation
• de-noising