Research On Fusion Of Fluorescent And Phase Contrast Images In Arabidopsis Thaliana Cells

Image fusion technique is an effective way to merge the information contained in different imaging modalities by generating a more informative composite image.In the field of cell and molecular biology,green fluorescent protein(GFP)images provide functional information embodying the molecular distribution of biological cells while phase contrast images maintain structural information with high resolution.Fusion of GFP and phase contrast images is of high significance to the study of subcellular localization,protein functional analysis,and genetic expression.Under the above circumstance,this dissertation concentrates on the study of multi-modality biological image fusion issue from the perspectives of transform domain and deep learning,and proposes three novel fusion methods.1.On the study of transform domain-based GFP and phase contrast image fusion methods,considering that traditional transform domain-based methods have limited ability in extracting detail information from the source images,a phase congruency and non-subsampled shearlet transform(NSST)-based GFP and phase contrast image fusion method is proposed.The input images are decomposed by NSST to get the high-frequency coefficients and the low-frequency coefficients first.Then,the high-frequency coefficients are fused with a strategy based on phase congruency and parameter-adaptive pulse coupled neural network(PA-PCNN),and the low-frequency coefficients are integrated through a local energy-based rule.Finally,the fused image is generated through conducting inverse NSST on the fused high-frequency coefficients and the low-frequency coefficients.Experimental results demonstrate that the proposed method can capture detail information from source images adequately,and outperform traditional methods.2.Considering that the traditional transform domain-based image fusion methods need manually design transformation methods and fusion strategies,and to fully take the characteristics of different input modalities into consideration,a GFP and phase contrast image fusion via generative adversarial networks(GANs)is proposed,in which the fusion problem is modelled as an adversarial game between a generator and a discriminator.The adversarial process enables that the fused result contain complementary information from the input modalities as much as possible.Experimental results demonstrate that the proposed method outperforms other representative GFP and phase contrast image fusion method on both objective evaluation and visual quality.3.Since the GAN-based fusion method has limited capability in capturing spatial detail information from the phase contrast image and the GAN model is sometimes unstable,a GFP and phase contrast image fusion method via detail preserving cross network(DPCN)is proposed.In this method,unlike traditional parallel multi-branch architectures used for multiple inputs,a structural-guided functional feature extraction branch and a functional-guided structural feature extraction branch are interacted via a cross manner to fuse the functional information from the GFP image and the structural information from the PC image more adequately.Moreover,the detail preserving module is composed of eight multi-scale convolutional blocks(MSCBs)associated with short,medium,and long skip connections to further extract the detail information from the source images.Experimental results demonstrate that the proposed DPCN-based method outperforms several state-of-the-art methods in terms of both visual quality and objective assessment.