A method for coding a sequence of digital images

Simple SummaryContent extracted from patent full text and abstract with AI.

This invention describes a method for coding (compressing) and decoding sequences of digital images using a novel approach to motion estimation. Unlike traditional block-based or vector-heavy pixel-based motion compensation methods, this technique represents motion between two images as a 'scalar field'—a smooth function from which the required motion vectors are derived through its gradient. This scalar field captures complex, deformable motion (like tissue movement in medical imaging) more efficiently and with less additional data, eliminating block artifacts common in conventional schemes.

Use CasesContent extracted from patent full text and abstract with AI.

• Compression of medical imaging sequences such as CT or MRI scans for storage or transmission.
• General video encoding and compression for streaming and broadcasting, where efficient handling of subtle or deformable motion is needed.
• Remote diagnosis and telemedicine applications requiring fast and efficient transmission of image sequences with high fidelity.
• Archival of large multiframe datasets in research or clinical healthcare environments.
• Encoding of multidimensional or volumetric scientific data, such as 3D or 4D simulations in research.

BenefitsContent extracted from patent full text and abstract with AI.

• Significantly reduces the amount of side information (metadata), improving compression rates compared to classic block-based or pixel-motion schemes.
• Removes block artifacts, resulting in higher quality images after decoding—especially important for high-accuracy applications like medical imaging.
• Enables precise modeling of smooth, complex, and deformable motion (such as soft tissue movement), which block-based methods cannot capture efficiently.
• Flexible and scalable: the number and location of parameters in the scalar field can be tuned according to performance or compression needs.
• Allows sub-pixel or sub-voxel motion accuracy without increasing the data overhead, improving prediction and reducing error residuals.
• Applicable not only for temporal (time-series) image sequences but also for spatial or volumetric (3D/4D) arrays.

Technical Classifications (CPCs)

Inventors & Applicants

Patent Abstract

The invention refers to a method for coding a sequence of digital images (S), each image having the same image format and comprising a number of pixels with assigned pixel values, wherein motion parameters (MP) between first and second images (I1, 12) are determined, where based on said motion parameters (MP) a motion compensation (MC) is performed for coding the sequence of images (S), where said motion parameters (MP) are included in the coded sequence of images (CS). The motion parameters (MP) between a first image (I1) and a second image (12) comprise a scalar field ( p ) having scalar values for a plurality of image positions in the image format, wherein the scalar field ( p ) is determined such that gradient vectors (MV) derived from the scalar field ( p ) correspond to motion vectors for the motion compensation (MC).