Advancements in medical image fusion: a comprehensive study on RTVD framework and FPGA deployment challenges
摘要
This paper presents the implementation of medical image fusion techniques based on Relative Total Variation Decomposition (RTVD) alongside appropriate software design and Field Programmable Gate Array (FPGA) utilization. RTVD is employed to decompose the images intended for fusion into their structural and texture components. The structural component primarily captures the large-scale structures and brightness of the source images, while the texture component focuses on finer details, including texture and noise characterized by low gradient values.Distinct fusion weights are created based on the characteristics of the structural and texture layers. To preserve texture information, the weights for the texture components are derived from a saliency map, whereas the weights for the structural components are determined based on image energy to maintain the brightness of the original images. Subsequently, the RTVD technique is applied to enhance the visual quality and information richness of the fused image. The paper also presents simulations and implementations of the RTVD-based fusion framework, comparing it with traditional techniques such as Principal Component Analysis (PCA), Additive Wavelet Transform (AWT), Discrete Wavelet Transform (DWT), Fourth-order Partial Differential Equations (FPDE), Multiscale Guided Filter-based Fusion (MGFF), and Multiscale Singular Value Decomposition (MSVD). The RTVD-based fusion technique is synthesized using the Spartan-3E FPGA Kit and Xilinx ISE Design Suite 14.5, and validated through MATLAB Simulink. Qualitative and quantitative experiments conducted on public datasets demonstrate the effectiveness of the RTVD method. The results indicate that the RTVD fusion method achieves high metric values (