Variational segmentation of vector-valued images with gradient vector flow.

Area of Science:

• Medical imaging
• Image processing
• Computer vision

Background:

• Variational segmentation methods are crucial for image analysis.
• Existing methods struggle with noise and accuracy in vector-valued images.
• Gradient vector flow (GVF) is a powerful tool but requires enhancement for complex data.

Purpose of the Study:

• To develop a robust variational segmentation method for vector-valued images.
• To improve the accuracy of gradient vector flow fields using enhanced edge information.
• To apply and validate the method for biological volume delineation in dynamic PET imaging.

Main Methods:

• Extended the gradient vector flow field using a vectorial edge map.
• Defined the edge map from a weighted local structure tensor.
• Incorporated a blind contrast estimator to weight image channels and reduce noise.
• Utilized the 4D gradient vector flow equation for accurate vector diffusion.

Main Results:

• The proposed method demonstrates robust variational segmentation of vector-valued images.
• Accurate diffusion of gradient vectors is achieved through the enhanced GVF field.
• Noise reduction is effectively managed by weighting image channels based on contrast.
• Successful application to biological volume delineation in dynamic PET imaging.

Conclusions:

• The enhanced GVF method provides robust and accurate segmentation for vector-valued images.
• The technique shows significant potential for applications in medical imaging, particularly PET.
• Validation on simulations and real images confirms the method's efficacy and reliability.