ISO/IEC 15938-15:2019

Overview

ISO/IEC 15938-15:2019 - Compact descriptors for video analysis (CDVA) specifies a standardized descriptor format and extraction/encoding process for visual content matching in video. The standard targets search and retrieval applications that require robust, compact video descriptors for matching object instances and scenes across large video databases (broadcast archives, internet video). It emphasizes small descriptor size, short runtimes, interoperability and efficient implementation on professional and embedded systems. The scope excludes person or face identification.

Key topics and requirements

• CDVA descriptor structure
  • Defines descriptor components including global descriptors, local descriptors, and deep feature descriptors for video analysis (see Clause 6).
  • Specifies a binary bitstream syntax and header/segment structures to enable interoperable exchange (see Clause 5).
• Extraction and encoding procedures
  • Normative steps for extracting and encoding CDVA descriptors to minimize size while preserving matching performance (Clause 6).
  • Annex A provides recommended parameter values for the encoding process.
• Deep feature extraction
  • Parameters and a neural network model for deep features are described (Annex B), enabling compact learned representations suitable for video.
• Robustness and performance goals
  • Designed for robustness to partial occlusion, viewpoint and camera changes, and varying lighting conditions.
  • Prioritizes accuracy vs. complexity trade-offs and reduced runtime compared with single-frame visual descriptors.
• Interoperability and implementation
  • Bitstream syntax, descriptor semantics and recommended parameters to support interoperable object instance search systems and efficient implementation on embedded and professional platforms.

Applications and users

• Video search & retrieval platforms
  • Large-scale content matching across broadcast archives, streaming libraries, and user-generated video repositories.
• Media asset management
  • Automated cataloging, duplicate detection and content-based retrieval for broadcasters and OTT services.
• Embedded and edge devices
  • Low-latency visual matching on cameras, mobile devices and edge analytics systems that require small, efficient descriptors.
• Research and product development
  • Developers of visual search engines, video analytics tools and computer vision systems seeking a standardized descriptor format.

Typical users: software engineers, system architects, multimedia indexing teams, researchers in computer vision and companies building video search, content monitoring or asset-management solutions.

Related standards

• ISO/IEC 15938-13:2015 - Compact descriptors for visual search (still-image descriptors); CDVA complements this for video.
• Other parts of the ISO/IEC 15938 (MPEG-7) series covering systems, description languages and profiles.