Image restoration model for microscopic defocused images based on blurring kernel guidance

Area of Science:

• Microscopy
• Image Processing
• Computational Imaging

Background:

• Defocus blur in microscopy degrades image quality and limits applications.
• Accurate blurring kernel estimation is crucial for high-resolution image restoration but computationally intensive.
• Existing neural network methods often require large datasets and have limited restoration resolution.

Purpose of the Study:

• To develop an efficient and accurate image restoration method for microscopic defocused images.
• To address the challenges of blurring kernel estimation in microscopic imaging.
• To enhance the observation accuracy and application range of optical microscopes.

Main Methods:

• A defocused image classification network was designed to handle varying defocus distances and directions.
• A blurring kernel extraction network (feature extraction, correlation, reconstruction) was developed.
• A non-blind restoration model based on U-Net integrated the blurring kernel extraction module, with joint training of kernel estimation and image restoration losses.

Main Results:

• The proposed method significantly improves peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM).
• Experimental results show superior performance compared to existing image restoration techniques.
• The method effectively restores high-resolution images from defocused microscopy.

Conclusions:

• The proposed blurring kernel estimation-guided image restoration method enhances microscopic image quality.
• This approach overcomes limitations of traditional methods regarding computational cost and dataset requirements.
• The developed technique offers a promising solution for improving optical microscopy observations.