ISO/IEC TR 29170-1: Information Technology — Advanced Image Coding — Part 1: Evaluation Framework

The landscape of image coding technology has expanded dramatically beyond the well-known JPEG standard. ISO/IEC TR 29170-1 establishes an evaluation framework for advanced image coding technologies, providing a standardized methodology for comparing coding efficiency, visual quality, and computational requirements across different codecs. This technical report is essential reading for engineers designing systems that must select or evaluate image compression technologies.

Evaluation Framework and Performance Metrics

The standard defines a comprehensive evaluation framework covering three domains: compression efficiency, visual quality, and resource requirements. Compression efficiency is measured through rate-distortion curves plotting decoded image quality (measured by PSNR, SSIM, or VMAF) against bitrate across a diverse test image set. The standard specifies a reference test set containing images with varying content types (natural scenes, text, computer graphics, medical imagery) at multiple resolutions.

Visual quality assessment in ISO/IEC TR 29170-1 combines objective metrics (PSNR, SSIM, MS-SSIM, VMAF, and CIEDE2000) with subjective testing methodologies derived from ITU-R BT.500. The standard acknowledges that objective metrics alone are insufficient for evaluating perceptual quality, particularly at low bitrates where different codecs exhibit different artifact characteristics — blocking, ringing, blurring, or color shifts — each perceived differently by human viewers.

Engineering Considerations for Codec Selection

Selecting an image coding technology for a production system involves trade-offs beyond compression efficiency. ISO/IEC TR 29170-1 provides guidance on evaluating computational complexity (encoding and decoding time, memory usage, power consumption on target hardware), encoding latency (critical for real-time applications), parallelization capability, and hardware acceleration availability. The standard introduces the concept of “operating points” — specific combinations of encoding parameters that achieve a target balance between compression ratio and computational cost.

Decoding complexity is particularly important for web and mobile applications where content is encoded once but decoded millions of times. A codec requiring 5x more decoding computation per image may be unacceptable even if it provides 20% better compression, when the additional decoding cost is multiplied across millions of users. The standard’s framework quantifies these trade-offs explicitly, enabling data-driven codec decisions.

Backward compatibility and ecosystem support are also addressed. The standard considers the availability of encoder/decoder implementations, maturity of software libraries, patent licensing implications, and browser/OS support. An advanced codec with superior compression but limited deployment reach may be impractical for consumer-facing applications, whereas server-side image processing pipelines can more readily adopt new technologies.

Future-Proofing Image Coding Architectures

ISO/IEC TR 29170-1 also discusses architectural strategies for systems that need to support multiple codec generations. Container formats supporting multiple codec types (such as HEIF and AVIF) enable graceful transitions between codecs as ecosystem support evolves. The standard recommends designing image processing pipelines with a codec abstraction layer that isolates application code from specific codec implementations, enabling codec upgrades without system-wide architectural changes.

An important aspect of the evaluation framework is the treatment of encoding time variability. Modern codecs often have substantially different encoding versus decoding complexity, and some use statistical multiplexing approaches where encoding time varies significantly between different images. For batch processing pipelines, this variability affects scheduling and throughput predictability. The standard recommends measuring encoding time histograms rather than simple averages to capture the full distribution of encoding performance under realistic operating conditions.

For streaming and real-time applications, the standard introduces latency-aware evaluation metrics that consider the encoding delay introduced by various codec tools. Intra-frame prediction modes, perceptual optimization passes, and rate control convergence all contribute to encoding latency, and each codec handles these differently. The evaluation framework provides specific test configurations for measuring end-to-end latency under different encoding parameter combinations.

Frequently Asked Questions