A Hybrid Denoising Algorithm of BM3D and KSVD for Gaussian Noise in DoFP Polarization Images

Abstract

In this paper, we present a hybrid denoising algorithm dedicated to division-of-focal plane (DoFP) polarization images. The proposed algorithm, centered around the Block-Matching and 3D Filtering (BM3D) and K-times Singular Value Decomposition (KSVD) denoising algorithms, is capable of significantly enhancing the grouping step in the second round of collaborative filtering by purifying the 'Semi-Filtered' image yielded by the first round of collaborative filtering. To achieve this, the BM3D denoising method's chain of operation is broken, and the 'Semi-Filtered' image is passed through a round of KSVD denoising method before the second round of collaborative filtering is conducted. According to our extensive experimental results, the proposed algorithm visually outperforms the state-of-the-art BM3D denoising algorithm and a wide range of other denoising algorithms for DoFP polarization images. Quantitative results presented using Peak-Signal-to-Noise-Ratio (PSNR) and Structural Similarity Index (SSIM) Index metrics further highlight the superior performance of the proposed algorithm.