Image and Video Compression Techniques

Image and video compression techniques aim to reduce the volume of data required to represent visual information while preserving perceptual quality. Broadly, methods fall into lossless approaches, which enable exact reconstruction, and lossy approaches, which tolerate controlled distortion for greater size reduction. Traditional lossless schemes rely on entropy coding, dictionary methods and predictive coding to exploit statistical redundancy. Lossy image codecs typically employ transform coding—such as discrete cosine or wavelet transforms—followed by quantisation and entropy coding. Video compression builds on intra-frame coding similar to still images, augmented by inter-frame prediction and motion compensation to exploit temporal redundancy. Hybrid block-based standards have dominated for decades, evolving from JPEG and MPEG to H.264/AVC and H.265/HEVC. Recent advances include versatile standards for emerging formats, adaptive streaming and immersive media, alongside entirely new paradigms based on deep neural networks. Neural methods use learned nonlinear transforms, variational autoencoders and entropy models to optimise rate–distortion performance under perceptual metrics. These developments have global significance in streaming services, remote sensing, medical imaging, satellite downlink and mobile communications, where efficient storage and real-time delivery of high-resolution content are essential.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Image and Video Compression Techniques publication trend

The graph below shows the total number of articles in image and video compression techniques across all publications each year (not limited to Nature Index journals).

Technical terms

Transform coding: A process that converts spatial image data into a frequency domain for more efficient quantisation and compression.

Entropy coding: A lossless technique that assigns shorter codes to more probable symbols, reducing redundancy.

Motion compensation: A video-coding tool that predicts blocks in one frame from shifted blocks in reference frames to exploit temporal correlation.

Rate–distortion trade-off: The balance between bit-rate reduction and the resultant quality loss in compressed data.

Autoencoder: A neural network architecture that learns to encode input data into a compact latent representation and decode it back for reconstruction.

Quantisation: The process of mapping continuous transform coefficients to a finite set of levels, enabling compression at the cost of precision.

References

1. Overview of the Versatile Video Coding (VVC) Standard and its Applications. IEEE Transactions on Circuits and Systems for Video Technology (2021).
2. Nonlinear Transform Coding. IEEE Journal of Selected Topics in Signal Processing (2020).
3. Intra Prediction and Mode Coding in VVC. IEEE Transactions on Circuits and Systems for Video Technology (2021).
4. Lossless Image Compression Techniques: A State-of-the-Art Survey. Symmetry (2019).
5. Subblock-Based Motion Derivation and Inter Prediction Refinement in the Versatile Video Coding Standard. IEEE Transactions on Circuits and Systems for Video Technology (2021).
6. Reduced-Complexity End-to-End Variational Autoencoder for on Board Satellite Image Compression. Remote Sensing (2021).