80 likes | 194 Views
Wavelet-based Image Fusion by Sitaram Bhagavathy Department of Electrical and Computer Engineering University of California, Santa Barbara
E N D
Wavelet-based Image Fusion by Sitaram Bhagavathy Department of Electrical and Computer Engineering University of California, Santa Barbara Source: “Multisensor Image Fusion using the Wavelet Transform,” by H. Li, B.S. Manjunath, and S.K. Mitra; Graphical Models and Image Processing, May 1999.
Outline • Objective: To integrate complementary information from multisensor image data such that the new images are more suitable for • perception, feature extraction, segmentation, object recognition, etc. • Wavelet-based fusion scheme: combines the DWTs of the input images and takes the inverse DWT Note: The input images have to be registered pixel-wise. • The basic algorithm • Modified feature selection algorithm • Results and conclusion Wavelet-based Image Fusion
The Basic Fusion Algorithm Wavelet-based Image Fusion
The Modified Algorithm • Activity measure: Maximum absolute value in a window centered at each pixel • Binary decision map created by maximum selection • IDWT after consistency verification Wavelet-based Image Fusion
Fusion of Grayscale images Output Input 1 Input 2 Note: I used 3 levels of decomposition, using the DB2 wavelet, for the experiments Wavelet-based Image Fusion
Fusion of Color Images I/P 1 I/P 2 Orig-inal O/P Wavelet-based Image Fusion
Conclusion • Wavelet-based fusion methods give better results than Laplacian pyramid-based methods • Fusion in the RGB color space works well but distorts the color at some pixels • Fusion in the YUV color space did not give good results; needs more experimentation Wavelet-based Image Fusion