Reliable and robust low rank representation based noisy images multi-focus image fusion

• Published in: Multimedia tools and applications 2023-03, Vol.82 (6), p.8235-8259
• Main Authors: Jagtap, Nalini, Thepade, Sudeep D.
• Format: Article
• Language: English
• Subjects: 1207: Innovations in Multimedia Information Processing & Retrieval; Coefficients; Color imagery; Computer Communication Networks; Computer Science; Computer vision; Data Structures and Information Theory; Discrete Wavelet Transform; Gray scale; Multimedia Information Systems; Noise reduction; Representations; Robustness; Special Purpose and Application-Based Systems; Wavelet transforms
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-021-11576-7

The noisy images fusion is still a challenging multi-focus image fusion (MIF) problem as the noise is inevitable for an input image. But most of the recent works do not bother about noisy images fusion and become challenging for color images. The fusion of noisy images pairs using existing MIF techniques compromises the robustness and reliability. We propose a novel framework to achieve the robust and reliable noisy images MIF called noisy image MIF (NIMIF). The NIMIF consists of the hybrid denoising technique, low-rank representation (LRR), and discrete wavelet transform (DWT). We propose two different NIMIF systems for greyscale MIF and color MIF. In color NIMIF, the input RGB color images have first converted into YCbCr color space due to their robustness. Then the Y color space image or greyscale image (in greyscale NIMIF) is decomposed using the DWT into high and low-level coefficients of each input image. We fuse the low-frequency coefficients using the spatial frequency (SF) technique. Before fusing the high-frequency coefficients, we apply the hybrid thresholding to suppress the noisy data from the input source images. The outcome of hybrid thresholding denoising fed to LRR to produce the fusion of high-frequency coefficients. Finally, NIMIF applies the inverse DWT to produce the MIF outcome. We present the comparative analysis of greyscale and color NIMIF using objective and visual results compare to state-of-art techniques. The simulation results prove that the integrated hybrid thresholding and LRR fusion technique form the reliable MIF solution for noisy greyscale and color image pairs.