Image Fusion with Enhanced NSCT Algorithms

Application Background:

This experiment is designed for MATLAB-based image fusion research. It performs multi-focus image fusion using various source images including "pepsi" and "clock" samples. The code has been modified with significant improvements to the high-frequency and low-frequency processing algorithms. Developed as part of my graduation thesis, the current implementation demonstrates robust stability and reliable performance.

Key Technologies:

The implementation utilizes image fusion through point-wise Non-Subsampled Contourlet Transform (NSCT) processing. The foundation is built upon the NSCT toolbox where low-frequency components are fused using the maximum pixel method, while high-frequency components employ the maximum variance method. The enhanced algorithm introduces pixel correlation-based fusion techniques, which analyze inter-pixel relationships to optimize fusion decisions. The code structure involves NSCT decomposition, coefficient analysis using variance calculations for high-frequency bands, and pixel-intensity comparisons for low-frequency reconstruction.