Image Encoding and Decoding Method and Device for Performing BDOF Sharper Motion in Compressed Video

This patent describes an image and video coding method that uses a technique called bi-directional optical flow, or BDOF, to refine how moving parts of a picture are predicted and reconstructed. It comes from Nokia, a major contributor to international communications and video standards.

The Problem

Video compression works largely by prediction. Rather than storing every frame in full, a codec predicts what a block of pixels will look like based on other frames, then stores only the small difference. For moving objects, the encoder estimates motion between frames and uses it to build the prediction. The catch is that motion estimation is done at a fairly coarse, block-by-block level, so the predicted picture is never quite right for objects that move smoothly or deform. Those small errors cost bits to correct and can leave visible flaws. Codec designers need a way to refine motion-based predictions more finely, without making the decoder so complex that it cannot run on ordinary phones and televisions.

What This Invention Does

The method determines whether bi-directional optical flow applies to the current block, and when it does, it refines the prediction at the sub-block level. It derives a prediction sample for the block from the block’s motion information, then, for a sub-block, it calculates a gradient, meaning how the pixel values change across the image. From that gradient it derives a small motion refinement, a pair of values describing a tiny adjustment to the motion. Combining the gradient and the refinement gives a BDOF offset, which is applied to produce a refined prediction sample for the block. In plain terms, the technique fine-tunes the predicted picture for moving content by looking at how the image changes within each small region, so the prediction is closer to the real frame and fewer bits are needed to correct it. The patent also covers the matching encoding side and a method for transmitting the resulting bitstream.

Key Features

• BDOF decision. The method first determines whether bi-directional optical flow applies to a block.
• Gradient-based refinement. It derives an image gradient for each sub-block to guide the adjustment.
• Motion refinement. A small motion correction is computed from the gradient.
• BDOF offset. The gradient and refinement combine into an offset that sharpens the prediction.
• Full pipeline. The patent covers decoding, encoding, and bitstream transmission.

Who Is Behind It

The applicant is Nokia Technologies Oy, the Finnish company’s unit that manages much of its patent portfolio and standards work. Nokia is a long-standing participant in developing video and communications standards. The named inventors are Nae Ri Park, Hyeong Moon Jang, Jung Hak Nam, and Seung Hwan Kim. This filing is a divisional application from an earlier related application.

Why It Matters

Video is the largest single category of internet traffic, so improvements that let a codec describe motion more accurately translate into better picture quality and lower bandwidth at very large scale. BDOF is part of modern standards such as Versatile Video Coding, and a method that performs it efficiently helps keep decoders fast enough for everyday devices. Protecting the technique in Australia supports the company’s standards and licensing interests in the local market.

Related Concepts

• Video coding format – the compression context the method operates in.
• Motion estimation – the prediction step BDOF refines.
• Optical flow – the concept the BDOF technique is built on.
• Versatile Video Coding – a modern standard that uses BDOF.
• Nokia – the company behind the application.

---

AU 2026201940 was published in the Australian Official Journal of Patents on 2 April 2026 and is open for public inspection. Patent applications represent inventions that are sought to be protected and do not necessarily reflect commercially available products.