Abstract
Wavelet-based image fusion techniques can generally be classified into three main categories: orthogonal, biorthogonal, and non-orthogonal wavelets. Although these wavelet types share certain common characteristics, each one has unique properties in terms of image decomposition and reconstruction. These differences influence the quality and characteristics of the fused image produced by each method.
This project focuses on comparing several image fusion approaches that utilize wavelets from these three categories. It also provides a theoretical analysis of the factors that contribute to variations in fusion results. In many cases, using only a wavelet transform for image fusion does not produce optimal results. However, improved performance can be achieved by combining wavelet transforms with traditional image fusion techniques such as the Intensity–Hue–Saturation (IHS) transform or the Principal Component Analysis (PCA) transform.
Therefore, this work also examines methods that enhance wavelet-based image fusion by integrating IHS or PCA transformations. Since substitution in IHS or PCA methods affects only a single component, integrating these techniques with wavelet transforms simplifies the fusion process and improves computational efficiency. In addition, this combination helps preserve important color information in the final fused image.
To evaluate the effectiveness of different approaches, IKONOS and QuickBird satellite images are used as test datasets. Seven wavelet-based fusion methods are analyzed, including orthogonal wavelet fusion with and without decimation, biorthogonal wavelet fusion with and without decimation, the ‘à trous’ wavelet method, and hybrid techniques combining wavelet transforms with IHS and PCA transformations.
The results are evaluated through graphical representation, visual inspection, and statistical analysis. The study shows that integrating wavelet transforms with other fusion techniques can improve the quality of fused images. These integrated approaches help reduce ringing and aliasing effects while producing smoother and more visually consistent images. The final fusion results are influenced by several factors, including the type of wavelet used, the presence or absence of decimation, and the number of wavelet decomposition levels.