Lagrange interpolation device for upsampling, method thereof, and computer-readable recording medium storing the method

Upsampling device for reducing color errors using weighted sum of Lagrange interpolation with left, right, and center filters

This technology is an interpolation device that calculates Lagrange interpolation values using three filters, each selecting 6 pixels from the left, right, and center regions of an upsampled image, and then calculates the value at the midpoint between pixels by summing them with weights of 0.25, 0.25, and 0.5.

Upsampling tasks such as deinterlacing or demosaicing have historically faced challenges in inferring values when the number of missing pixels exceeds the actual data. Conventional interpolation methods often fail to adequately account for edge information, leading to color errors where unnatural colors appear at object contours or boundaries.

This technology places the greatest weight on the center region value among the Lagrange interpolation results of the three overlapping regions, ensuring stable color restoration at boundaries without being skewed by outliers in one region. It can be applied to image signal processing (ISP) chips in digital cameras and dash cams, suppressing color bleeding at boundaries through simple polynomial operations without increasing hardware load.

‍

‍

Key Features:

• Interleaving configuration that fills empty pixels with 0 after downsampling and then upsampling the image from the color filter array sensor
• Left, right, and center upsampling filters that each select 6 pixels from the left region, right region, and the center region overlapping both, on a Cartesian coordinate plane
• Channel interpolation unit that calculates first, second, and third pixel values at the midpoint between pixels by applying Lagrange interpolation to the pixels extracted by each filter
• Configuration that calculates the final pixel value by multiplying the first and second pixel values by 0.25 and the third pixel value of the center region by 0.5

‍