|
The fusion is a combination of sensors, image processing, signal processing, display, modern high-tech computer and artificial intelligence technology, the integration of multiple images of the same scene information technology, one on the same scene more clearly, contains more useful information about the image. Image fusion technology plays an important role in the field of medicine, surveying, geographic information systems, industrial, intelligent robots, and military, and increasingly the attention of many researchers. In the transform domain, wavelet-based image fusion method is used more often. The wavelet transform has a significant advantage in the one-dimensional signal processing, wavelet transform natural image can not make full use of the data itself unique geometric features can not be that high-dimensional function containing a line or plane singular optimal, at the same time wavelet transform capture directional information is also limited. Multiscale geometric analysis is the best representation of high-dimensional functions developed in recent years, mainly ridgelet transform, Curvelet transform, Bandelet transform, and Contourlet Transform and Nonsubsampled Contourlet Transform. Multiscale geometric analysis for pixel-level image fusion is proposed, based on in-depth study and discussion of image fusion rules and implementation methods, based on Nonsubsampled Contourlet transform and PCNN image fusion algorithm combined. The algorithm uses the spatial frequency and orientation contrast are to trigger PCNN, are better able to extract the source image characteristics coefficient effectively preserve the texture detail of the image, and greatly improve the fusion effect. Specifically, the paper detailed analysis Nonsubsampled Contourlet Transform, Contourlet transform and wavelet-Contourlet transform image fusion. Focus on research-based nonsubsampled Contourlet transform image fusion, and the introduction of PCNN image fusion field, combined with a new image fusion algorithm nonsubsampled Contourlet transform and PCNN. Through the the multifocus image and infrared remote sensing image fusion simulation, and verify the effectiveness of the algorithm. Image fusion results of evaluation of the field of image fusion is an important issue. A comprehensive analysis of existing image fusion of subjective and objective evaluation criteria, objective evaluation parameters (such as clarity, spatial frequency, mutual information, entropy, etc.) combined with subjective visual analysis of the simulation results. Through comparative analysis, non-sampling compared to Contourlet transform and other transforms, fusion better. NSCT-PCNN-based fusion algorithm greatly improves the overall performance of the image edge, texture, and retain more information of the source image, it is a universal strong and efficient image fusion method.
|