Multi-focus Image Fusion: Wavelet Decomposition and PSF Model Integration

In this paper, we conduct an in-depth discussion on multi-focus image fusion. Initially, we introduce the wavelet transform fusion algorithm. Although wavelet transform methods demonstrate advantages in non-redundancy and minimal high-frequency component loss, leading to satisfactory fusion performance, they exhibit two primary drawbacks: shift-variance characteristics and susceptibility to external factors during image reconstruction, resulting in partial edge information loss. To address these limitations, we implement a fusion methodology combining wavelet decomposition with Point Spread Function (PSF) modeling. The technical implementation begins with non-downsampled wavelet decomposition applied to different source images, preserving original dimensions through careful coefficient handling. Subsequently, we superimpose high-frequency components across various orientations and scales, extracting distinctive features to differentiate between focused and defocused targets through threshold-based classification algorithms. Based on this analysis, we design a fusion algorithm employing strategic weighting schemes to combine the two source images. By circumventing inverse wavelet transformations through direct coefficient manipulation, the fused image achieves enhanced visual quality with superior edge preservation and reduced artifacts compared to conventional wavelet-based approaches.