Limits...
Registration and fusion of the autofluorescent and infrared retinal images.

Kolar R, Kubecka L, Jan J - Int J Biomed Imaging (2008)

Bottom Line: This article deals with registration and fusion of multimodal opththalmologic images obtained by means of a laser scanning device (Heidelberg retina angiograph).The registration framework has been designed and tested for combination of autofluorescent and infrared images.Two fusion methods are presented and compared.

View Article: PubMed Central - PubMed

Affiliation: Department of Biomedical Engineering, Faculty of Electrical Engineering and Communication, Brno University of Technology, Kolejni 4, 61200 Brno, Czech Republic. kolarr@feec.vutbr.cz

ABSTRACT
This article deals with registration and fusion of multimodal opththalmologic images obtained by means of a laser scanning device (Heidelberg retina angiograph). The registration framework has been designed and tested for combination of autofluorescent and infrared images. This process is a necessary step for consecutive pixel level fusion and analysis utilizing information from both modalities. Two fusion methods are presented and compared.

No MeSH data available.


Effect of anisotropic diffusion filtering on (a) AF and (b) IR images. The corresponding images are on Figure 1. The parametersof anisotropic diffusion were set to κ = 1.5, t = 0.125, and iteration = 10.
© Copyright Policy - open-access
Related In: Results  -  Collection


getmorefigures.php?uid=PMC2570800&req=5

fig3: Effect of anisotropic diffusion filtering on (a) AF and (b) IR images. The corresponding images are on Figure 1. The parametersof anisotropic diffusion were set to κ = 1.5, t = 0.125, and iteration = 10.

Mentions: The first stepis the image preprocessing. Because of the low signal-to-noise ratio (SNR),noise suppression method should be performed, but possibly without blurring theedges which carry information needed for registration. For that reason, we usedthe anisotropic diffusion introduced by Perona and Malik [26]. The approach of this method is that a Gaussiansmoothed image is considered as a single time slice of the solution to the heatequation: (2)∂g(x, y, t)∂t = ∇⋅c(/∇g/)∇g(x, y, t), where g(x, y, 0) = f(x, y) is the input image. The function c(/∇g/) is a function that reduces the conductance atthe areas of large gradient /∇g/ . We used the following form [27]: (3)c(/∇g/) = e−/∇g/2/2κ2, which introduces a newconductance parameter κ, controlling sensitivity of the smoothing process. The behavior of this functionfor tested values of κ is depicted on Figure 2. The second parameter of anisotropic diffusion is time t and the last parameter is the number ofiterations. For the IR and AF images, we found as suitable values for κ = 1.5, t = 0.125, and iteration = 10. For higher values of κ, the smoothing effect was too high and the resulting images were not suitablefor consequential gradient computation (see Section3.2). The effect of this filter is shown on Figure 3.


Registration and fusion of the autofluorescent and infrared retinal images.

Kolar R, Kubecka L, Jan J - Int J Biomed Imaging (2008)

Effect of anisotropic diffusion filtering on (a) AF and (b) IR images. The corresponding images are on Figure 1. The parametersof anisotropic diffusion were set to κ = 1.5, t = 0.125, and iteration = 10.
© Copyright Policy - open-access
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC2570800&req=5

fig3: Effect of anisotropic diffusion filtering on (a) AF and (b) IR images. The corresponding images are on Figure 1. The parametersof anisotropic diffusion were set to κ = 1.5, t = 0.125, and iteration = 10.
Mentions: The first stepis the image preprocessing. Because of the low signal-to-noise ratio (SNR),noise suppression method should be performed, but possibly without blurring theedges which carry information needed for registration. For that reason, we usedthe anisotropic diffusion introduced by Perona and Malik [26]. The approach of this method is that a Gaussiansmoothed image is considered as a single time slice of the solution to the heatequation: (2)∂g(x, y, t)∂t = ∇⋅c(/∇g/)∇g(x, y, t), where g(x, y, 0) = f(x, y) is the input image. The function c(/∇g/) is a function that reduces the conductance atthe areas of large gradient /∇g/ . We used the following form [27]: (3)c(/∇g/) = e−/∇g/2/2κ2, which introduces a newconductance parameter κ, controlling sensitivity of the smoothing process. The behavior of this functionfor tested values of κ is depicted on Figure 2. The second parameter of anisotropic diffusion is time t and the last parameter is the number ofiterations. For the IR and AF images, we found as suitable values for κ = 1.5, t = 0.125, and iteration = 10. For higher values of κ, the smoothing effect was too high and the resulting images were not suitablefor consequential gradient computation (see Section3.2). The effect of this filter is shown on Figure 3.

Bottom Line: This article deals with registration and fusion of multimodal opththalmologic images obtained by means of a laser scanning device (Heidelberg retina angiograph).The registration framework has been designed and tested for combination of autofluorescent and infrared images.Two fusion methods are presented and compared.

View Article: PubMed Central - PubMed

Affiliation: Department of Biomedical Engineering, Faculty of Electrical Engineering and Communication, Brno University of Technology, Kolejni 4, 61200 Brno, Czech Republic. kolarr@feec.vutbr.cz

ABSTRACT
This article deals with registration and fusion of multimodal opththalmologic images obtained by means of a laser scanning device (Heidelberg retina angiograph). The registration framework has been designed and tested for combination of autofluorescent and infrared images. This process is a necessary step for consecutive pixel level fusion and analysis utilizing information from both modalities. Two fusion methods are presented and compared.

No MeSH data available.