Methods and devices for pixel-prediction for compression of visual data

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

This patent introduces improved methods and devices for predicting pixel values during the compression of visual data, such as images or videos. Unlike traditional techniques, it uses multiple model functions—static and adaptive—that are chosen based on how well they fit the characteristics of reference pixels. This leads to more accurate predictions, particularly for challenging regions like edges or areas with distinct intensity changes, resulting in better compression performance.

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

• Medical image compression (e.g., CT, MRI, or ultrasound images) for efficient storage and transmission
• General lossless and lossy image compression for digital photography, graphic design, or scientific imaging
• Video compression for applications that require high fidelity, such as telemedicine or surveillance
• Defective pixel interpolation for repairing corrupted image areas
• Error concealment and noise removal in image or video streams
• Super-resolution processing for enhancing image detail
• Region-of-interest coding for focusing on critical areas, e.g., in diagnostics or security
• Compression in hyperspectral imaging used in remote sensing or agriculture

BenefitsContent extracted from patent full text and abstract with AI.

• More precise pixel prediction, especially in images with sharp edges or complex structures
• Improved compression ratios, meaning reduced storage space and transmission bandwidth
• Enhanced image quality, particularly important for applications where lossless or near-lossless compression is required, such as medical diagnostics
• Adaptive modeling allows handling a wide variety of image characteristics, from smooth areas to detailed edges
• Reduced dependency on large buffer sizes or extensive context, making the method suitable for small images or incremental/image boundary regions
• Supports multiple dimensionalities—2D images, 3D volumes, hyperspectral data, and video
• Facilitates parallel processing implementations, improving processing speed
• Flexible integration with existing entropy coding and transform coding techniques

Technical Classifications (CPCs)

Inventors & Applicants

Patent Abstract

The invention relates to pixel-prediction methods and devices that are based on one of a selection of one out of at least two model functions that provide a prediction function and an adaptive model function that is capable to predict intensity characteristics of reference pixels used for prediction.