Autocorrelation-Based Interlace-to-Progressive Format Conversion

Interlace-to-Progressive Conversion Based on Edge Direction Determination Using Interpolation Error Autocorrelation

This technology is an interpolation method that calculates costs for candidate directions based on the autocorrelation of interpolation errors generated by predicting neighboring pixels around missing pixels. It identifies the direction with the lowest cost as the edge and fills the missing pixels using a 6-tap Blackman-Harris windowed sinc filter.

As video data surges with the spread of UHDTV and 3DTV, interlaced formats remain in use; however, since many display devices cannot display them directly, deinterlacing has become essential. Existing edge-based interpolation methods suffer from limited edge direction options, inefficient early termination, and inaccurate edge detection, leading to jagged diagonal boundaries.

By statistically comparing error patterns of neighboring pixels, this technology reliably detects edges while allowing for an arbitrary number of candidate directions. It combines the determined edge direction with BH windowed sinc predictions in the vertical direction to produce smooth interpolation results while preserving boundaries. Applicable to UHD TV upscaling engines and video archive restoration software, it reproduces images without aliasing, even in content with many diagonal and curved contours.

‍

‍

Key Features:

• Calculating interpolation errors for candidate directions for pixels neighboring a missing pixel in a low-resolution image
• Calculating the autocorrelation for each edge direction using the calculated interpolation errors
• Determining the edge direction by selecting the direction with the lowest cost, derived from the mean and variance of the interpolation errors
• Interpolating the missing pixel using a 6-tap weighted average filter, which is a BH windowed sinc function in the frequency domain

‍