Image processing method based on entropy-weighted allocation

High-Definition Interlace-to-Progressive Conversion Method Based on Information Entropy Weighting

This technology is a deinterlacing image processing method that restores missing lines in an interlaced signal to convert it into a progressive frame. It calculates pixel values using both vertical line averaging and diagonal edge-based methods, then determines the final pixel value by applying weights to these results using an information entropy map.

While deinterlacing that utilizes temporal information offers high performance, it requires motion modeling and tracking, leading to high computational complexity. Spatial domain methods, such as line recovery (Bob) or edge-based line averaging, are simpler but suffer from reduced image quality around edges, creating a need for more effective conversion methods.

This technology performs interpolation tailored to regional characteristics by selecting the averaging method if the entropy value of each target pixel exceeds a threshold between 0 and 1, and the diagonal edge method if it is below that threshold. As a result, it produces progressive video with improved sharpness compared to conventional Bob and EELA methods. It can be applied to video format conversion in broadcast receivers, TVs, and display driver chips, providing clear images with sharp edges without the need for motion estimation.

‍

‍

Key Features:

• Calculating a first pixel value for each target pixel using the average of vertical line pixel values from the interlaced signal
• Calculating a second pixel value for the target pixel using neighboring pixel values from the vertical lines and neighboring pixel values from different diagonal directions
• Generating an entropy map with values between 0 and 1 by calculating an entropy pixel value based on the probability that the target pixel is the actual pixel value
• Determining the final pixel value by selecting the first method if the entropy map value is greater than a threshold between 0.1 and 0.7, or the second method if it is below that threshold

‍