Multi-focus image fusion using Stationary Wavelet Transform (SWT) with Principal Component Analysis (PCA)

摘要

Multi-focus image fusion creates meaningful image from two or more meaningless images which have same scenes with meaningful image. These images have different focus points. The image after proposed method is named as all-in-focus image. This image has more information from source images. Multi-focus image fusion is that combining two or more source images which have same scenes but different focuses. In this paper, we proposed lifting wavelet transform based hybrid technique. Principal Component Analysis is used as a fusion rule. Firstly, source images are decomposed using Lifting Wavelet Transform. After this, all source images divided into four sub-bands. Secondly, the each sub-band of source images is applied Prinicipal Component Anlaysis. And eigenvectors and eigenvalues are calculated. Calculated eigenvectors are used to fuse sub-bands. Finaly, the new sub-bands are created and Inverse Lifting Wavelet Transform is implemented for new sub-bands. The fused image is created and to perform quality Mutual Information, Petrovics metric and Average Gradient are calculated. The results show that the new hybrid technique is successful for multi-focus image fusion. All in focus image is more informative so it can be processed easily. Multi-focus image fusion is used different areas such as; health system, wsn, etc. We proposed a new hybrid method using Stationary Wavelet Transform (SWT) with Principal Component Analysis (PCA). This method uses transform domain. We used SWT for feature extraction. SWT decompose image four different sub-bands. After extraction feature, to combine images we proposed PCA based fusion rule. With PCA from sub-bands of source images are computed eigenvectors and selected maximum eigenvector of these sub-bands because maximum eigenvector represents image ideally. After application fusion rule, we got four new sub-bands and reconstructed new all in focus image using this sub-bands with inverse SWT. Mutual Information, Standard Deviation, Spatial Frequency and Petrovic's Metric are used as quality metrics.