In this blog post, we’ll take an easy-to-understand look at the principles of deinterlacing—the process used to enable seamless playback of high-definition digital video—its key algorithms, and how it’s applied in today’s video technology.
Deinterlacing and Its Background
Today, we use smartphones to listen to music, watch movies, make video calls with friends, and interact in real time with people in other regions through virtual reality (VR) and augmented reality (AR) environments. These experiences are made possible by high-speed communication and signal processing technologies. The field of communications focuses on researching methods to efficiently transmit and receive information, while the field of signal processing focuses on technologies to improve transmission efficiency and restore received signals to be as close to the original as possible. De-interlacing is a key technology in the field of video restoration within signal processing.
To transmit high-definition video in real time, a very large amount of data must be transmitted quickly. The reason video data is so large is that each frame consists of countless pixels. A pixel, short for “picture element,” is the smallest unit that makes up a display. Generally, a single pixel represents color using three color channels—red (R), green (G), and blue (B)—with each channel typically using a value of 8 bits (0–255). Therefore, a single pixel can represent approximately 16.77 million (256³) different colors, and as resolution increases, the amount of data required to display a single frame also increases significantly.
To address this issue of increasing data volume, past broadcasting systems utilized the interlacing method. Interlacing is a method that transmits odd and even lines alternately rather than transmitting the entire screen at once; it was effective for improving video transmission efficiency under the limited bandwidth conditions of the time. Interlaced video transmitted in this way must be restored at the receiving end to look as close to the original as possible, and this restoration process is known as deinterlacing.
Today, while most digital broadcasting and internet streaming have switched to the progressive scan method, interlaced video is still used in legacy broadcast content and some video equipment. Therefore, deinterlacing technology is not a relic of the past but remains a vital video processing technique used today to ensure the natural playback of existing content.
While the interlaced format has the advantage of reducing bandwidth requirements, playing it back as-is can result in visual artifacts such as staircase effects, screen flicker, and motion distortion. Therefore, on the receiving end, post-processing techniques such as interpolation, averaging, and motion analysis are performed to compensate for missing information and produce a more natural video. Next, we will examine these representative deinterlacing algorithms.
Deinterlacing Algorithms and Practical Applications
Linear Average (LA) and Efficient Linear Average (ELA)
The most intuitive and widely known deinterlacing method is the Linear Average (LA). This method fills in the missing line’s pixel values by calculating the average of the surrounding pixels. There are two main methods for calculating this average. One is spatial averaging, which averages the values of the top and bottom lines within the same frame, and the other is temporal averaging, which uses pixels at the same position in the previous and next frames.
When there is almost no motion in the video, temporal averaging can produce results similar to or even better than spatial averaging. However, in fast-moving scenes, caution is required because using information from both the previous and next frames can cause ghosting or blurring.
Efficient Linear Average (ELA) is an improved algorithm designed to reduce the computational load of LA while maintaining a similar level of restoration quality. The basic concept is to reduce unnecessary computations and utilize the relationships between neighboring pixels more efficiently. However, since both LA and ELA rely on averaging for restoration, while they restore horizontal edges relatively well, they often produce stair-step artifacts (aliasing) along diagonal edges, such as the boundary lines on a ping-pong table.
Directional Interpolation (DOI) for Edge Correction and Its Applications
The Direction-Oriented Interpolation (DOI) family of algorithms was developed to reduce this stair-step effect. DOI analyzes the boundary direction of neighboring pixels and then performs interpolation in the same direction as the boundary. In other words, in areas with a gradient, it interpolates information along that direction to restore boundaries that are more natural and closer to the original.
The DOI method provides significantly better visual quality than LA or ELA, particularly in images with many diagonal or curved edges. Although its computational complexity is somewhat higher than that of LA, recent significant improvements in the performance of mobile devices, TVs, graphics processing units (GPUs), and dedicated video processing hardware have led to its widespread use in real-time applications. Furthermore, when combined with motion detection technology, it can effectively preserve the edges of moving objects, making it widely utilized in actual video processing systems.
Deinterlacing technology is used in various fields, including digital TV, broadcasting equipment, video editing systems, certain streaming services, video calls, medical imaging, and industrial video equipment. Furthermore, similar concepts are widely applied not only to video but also to the processing of one-dimensional signals, such as audio, to reduce data transmission volume, minimize quality degradation at the receiving end, and restore the signal.
Ultimately, the direction of video transmission technology can be broadly divided into two categories. One involves advancing network and wireless communication technologies to transmit more data per unit of time, while the other utilizes efficient compression and signal processing technologies to transmit less data while still restoring the signal at the receiving end to a quality similar to the original. Today, as high-efficiency video compression technology and AI-based video restoration technology advance in tandem, these two approaches are being used complementarily. Deinterlacing has also established itself as a key component of these video restoration technologies, and in real-world environments, appropriate algorithms are selected by comprehensively considering factors such as bandwidth, latency, and computational performance.
As information and communication technology advances, people expect ever-sharper picture quality and more natural video playback. These demands, in turn, are driving the development of new signal processing technologies, video compression technologies, and AI-based video restoration technologies. Technologies like deinterlacing—which are difficult for users to perceive directly—significantly influence the final viewing experience and are expected to continue evolving in conjunction with various video processing technologies.
