Maximum induction-based intra-field deinterlacing method

Intra-field deinterlacing based on hybrid estimation combining point and patch estimators

This technology is a maximum induction-based method that performs deinterlacing by calculating pixel-wise point estimators and vectorized patch-wise estimators for interlaced low-resolution images, then linearly combining them into a hybrid estimator.

Interlaced signals cause edge flickering, line flickering, and field aliasing on high-resolution displays, and because ideal sampling filters cannot be implemented in real systems, it has been difficult to fully restore original frequency components. Existing intra-field methods relied on either pixel-level or region-level estimation, leading to image quality degradation due to the inherent limitations of each approach.

This technology combines point-wise and patch-wise approaches into a hybrid estimator with regularization elements and models the estimation error distribution as Gaussian or Laplacian to interpolate according to image characteristics. The two methods compensate for each other's weaknesses, suppressing aliasing while maintaining computational efficiency. It can be applied to progressive scan conversion modules in HDTVs, PC monitors, and video upscaling software, allowing for the restoration of near-original progressive images using only single-field information.

‍

‍

Key Features:

• Calculating a point-wise estimator for an interlaced image that reflects the estimation error between the low-resolution image and its reconstructed version
• Calculating a patch-wise estimator using vectorized pixel segments for the interlaced image
• Calculating a hybrid estimator containing regularization elements based on the point-wise and patch-wise estimators, and performing deinterlacing using it
• Modeling the hybrid estimator as a Gaussian or Laplacian distribution with scale factors and normalization elements

‍