Optimized VLSI Implementation for Visible and Infrared Image Fusion using Stationary Wavelet Transform

Abstract

Image fusion has long relied on conventional signal processing techniques like discrete wavelet transform (DWT), contourlet transform, shift-invariant shearlet transform, and quaternion wavelet transform. However, these methods can introduce artifacts into the fused image, leading to suboptimal results. To address these issues, optimization-based fusion schemes have been proposed, although they often require multiple iterations to find the optimal solution, potentially resulting in oversmoothed images. This research focuses on a hardware-oriented VLSI-based implementation of visible-infrared (VI-IR) image fusion. The process begins by reading images in the MATLAB environment and applying stationary wavelet transform to decompose VI and IR images into multiple bands. Low-low bands are converted into text files, and a band fusion rule is applied using a multiplexer-based adder to combine both images. The final fused image is reconstructed in MATLAB. Simulation results demonstrate that this proposed method offers improved performance.