Spiking Cortical Model For Image Fusion

The pulse coupled neural network (PCNN) is known as the third generation of artificial neural network theory which originated from the study of mammalian visual cortex neurons. PCNN has been successfully used in image processing due to its visual characteristics of mammals. Spiking cortical model (SCM) is one of the simplified models of PCNN; it is closer to biological properties of visual neurons and has lighter computation comparing with traditional PCNN.The research work of this thesis mainly discusses SCM based image fusion technology in multi-focus image field and multi-sensor image field. The main work and achievements of the thesis are:1. Research on SCM based multi-focus image fusion technology. In this part, we proposed a new setting method of the iteration number of neural networks cycles; we proposed a new pixel resolution evaluation criteria and verified its validity, and then the algorithm framework and step of SCM based multi-focus image fusion are provided.2. Research on multi-sensor medical image fusion that based on SCM and nonsubsampled contourlet transform (NSCT). In this part, we firstly summarized the progress of existing image fusion technologies that based on PCNN and NSCT. Since NSCT is regarded as the representative method of existing multi-scale analysis methods, we combined SCM with NSCT in image fusion method. The SCM’s human visual characteristic is considered while designing fusion rule; after that, the fusion frameworks and algorithm steps are provided.3. Research on multi-source image fusion method based on SCM and discrete wavelet transform (DWT). Since DWT is considered as the most classic, sophisticated multi-scale decomposition algorithm, in this part we combined SCM with DWT in image fusion. After image is decomposed by NSCT, we use SCM to select high frequency sub-band coefficients of image; at the same time, we provided the detailed fusion steps of the proposed fusion method.The research work presents theoretical basis, algorithm framework, and the corresponding experiments for three proposed fusion method. In order to prove the effectiveness of the proposed methods, we provide both objective quantitative analysis and subjective qualitative analysis. The experimental results and comparative analysis verified the effectiveness of the proposed fusion method. Thus, the fusion methods of our research have some guidance and reference to solve image fusion problems.