Edge-Guided Image Gap Interpolation Using Multi-Scale Transformation

Area of Science:

• Computer Vision
• Image Processing
• Signal Processing

Background:

• Image gap restoration is crucial for reconstructing corrupted or missing image data.
• Existing methods often struggle with preserving both local textures and global object edges.

Purpose of the Study:

• To introduce an improved image gap restoration technique using edge-based directional interpolation.
• To enhance the accuracy and quality of reconstructed images, especially at high data loss rates.

Main Methods:

• Incorporation of edge-based directional interpolation within multi-scale pyramid transforms.
• Reconstruction of local edges (textures) and global edges (object boundaries) using gradient inference and Canny detection.
• Iterative estimation and up-sampling of missing samples across pyramid layers.

Main Results:

• The proposed method demonstrates significant improvements in peak-signal-to-noise-ratio (PSNR), ranging from 1-5 dB.
• Effective restoration across various image types with regular and random loss patterns up to 40% loss rates.
• Outperforms existing state-of-the-art image restoration techniques.

Conclusions:

• Edge-guided interpolation within multi-scale pyramid transforms offers superior image gap restoration.
• The method effectively preserves image details and structural integrity even with substantial data loss.
• Provides a robust solution for challenging image inpainting and restoration tasks.