Deinterlacing device and method based on bilinear and fuzzy-based weighted average filters

Deinterlacing using bilinear and fuzzy weighted average filters based on window variance

This technology is a device and method for generating deinterlaced images by initializing missing pixels in interlaced video using bilinear interpolation, estimating window variance to classify windows into heterogeneous or homogeneous types, and applying a fuzzy-based weighted average filter to heterogeneous windows and a bilinear filter to homogeneous windows.

HDTV broadcasting uses interlaced formats like 1080i, which transmit only half the lines per frame, but modern LCD and PDP displays are progressive, making conversion essential. Previous methods struggled to distinguish between flat and edge regions, often resulting in damaged edge detail or slow processing speeds due to complex computations.

This technology ensures both high image quality and processing efficiency by using a lightweight bilinear filter for flat areas and a fuzzy filter—integrating intensity similarity and geometric proximity weights—only for edge-heavy areas. By using a window variance threshold of 0.062 for classification, the decision process is simplified. It can be applied to digital TVs, set-top boxes, and broadcast conversion equipment to display interlaced signals clearly on progressive screens in real time.

‍

‍

Key Features:

• A missing pixel initialization unit that initializes missing pixels in an interlaced image using bilinear interpolation
• A variance estimation unit that estimates the variance of a given window and classifies the window as either a heterogeneous or homogeneous window based on the estimated variance
• A fuzzy-based weighted average filter that processes heterogeneous windows by integrating weights from intensity similarity and geometric proximity filters
• A deinterlaced image formation unit that applies bilinear filtering to homogeneous windows and forms a deinterlaced image based on the outputs of the two filters

‍