Video Coding in India: Smarter Compression for Streaming

The problems being solved

Video coding innovation in India homes in on the stubborn bottlenecks that inflate bitrates and degrade quality. One cluster of work targets intra prediction signaling—how a block guesses its content from already-decoded neighbors. Getting the most probable mode (MPM) list right, deriving chroma modes from luma, and fusing multiple derivation methods like DIMD and TIMD are all active fronts. The goal is to shrink the overhead of telling the decoder which prediction direction to use.

Motion estimation and prediction form another battleground. Here, the challenge is to improve accuracy without ballooning the side information for motion vectors. Innovators are rethinking merge candidate lists, reordering history-based candidates by template matching cost, and refining bidirectional optical flow (BDOF) offsets. The same thread runs through inter-intra mixing, where blending predictions from different references demands careful refinement.

In-loop filtering—the stage that cleans up artifacts after reconstruction—is being reimagined to handle modern content. Problems include how to pad unavailable samples for cross-component adaptive loop filtering (CC-ALF), how to make adaptive rounding more content-aware, and how to design nonlinear ALF functions that look at neighboring samples. Sample adaptive offset (SAO) classification and deblocking parameter signaling are also being tightened.

Syntax signaling itself is a drag on compression. Conditional encoding of secondary transform flags, independent signaling of filter enable flags, and smarter context model selection for tool flags are all ways inventors are cutting the fat. Palette mode signaling and quantization parameter handling are being streamlined, often by tying them to skip mode or other already-known conditions.

Finally, sub-picture and boundary handling addresses the messy edges of tiled and segmented video. Two-stage padding for boundary blocks, conformance cropping window conditions, and sub-bitstream extraction rules ensure that modern codecs can cleanly split a picture without breaking prediction or filtering across tile boundaries.

How the field is solving it

The technical approaches emerging from India are pragmatic and deeply tied to the latest codec standards like VVC. For intra prediction, a recurring theme is smarter MPM list construction—deriving modes from neighboring blocks, applying size-based restrictions, and conditionally parsing the MPM index so that only the most likely modes consume bits. Some solutions fuse decoder-side intra mode derivation (DIMD) with template-based intra mode derivation (TIMD) into a single hybrid mode, letting the encoder pick the best fusion on the fly.

Motion estimation is being sharpened through cost-based reordering. Instead of a fixed candidate order, history-based motion vector predictor (HMVP) lists are re-sorted using template matching costs, putting the most promising candidates first. For refinement, gradient-based BDOF offsets are derived with higher precision, and motion estimation itself is being recast using first, second, and third differences to capture subtle motion.

Boundary and padding techniques tackle the real-world complexity of sub-pictures. Two-stage padding—first motion-compensated padding, then repetitive padding—ensures that blocks at picture edges have valid reference data. Mirrored padding for unavailable luma samples in CC-ALF prevents filter discontinuities. These methods are baked into the parsing and conformance rules so that sub-bitstream extraction remains standards-compliant.

Hybrid and combined prediction modes are another vector. Beyond fusing DIMD and TIMD, inventors are exploring inter-intra mixing as a distinct coding mode, where the final prediction is a weighted blend of inter and intra signals. This sits alongside adaptive filtering innovations: nonlinear ALF functions that use neighboring sample differences, cross-component classifiers for SAO, and adaptive rounding in loop filters that adjusts based on local statistics.

Across all these, a common thread is conditional signaling—tying the presence of syntax elements to already-decoded flags or block types. This reduces the bitstream overhead without sacrificing flexibility, a hallmark of the efficiency-first mindset driving the work.

Where the market is heading

The global advanced video coding market was valued at roughly USD 4-5 billion in 2024, according to SNS Insider, and is growing at a mid-single-digit annual rate (SNS Insider, Research and Markets). The primary engine is the insatiable appetite for streaming video. OTT platforms are pushing 4K and 8K content, while live sports and events demand real-time, high-quality encoding. Video conferencing, now embedded in business, education, and healthcare, adds another layer of urgency for low-latency, bandwidth-efficient codecs.

India mirrors these global currents with its own explosive data consumption. Video already dominates mobile traffic, and the shift toward short-form content, regional OTT, and interactive live streams is accelerating. The country’s deep pool of engineering talent is responding with innovations that target the specific pain points of next-generation codecs—exactly the kind of IP that becomes critical as device makers and service providers look to squeeze more quality out of every bit.

Immersive formats like VR and 360-degree video are still niche but growing, and they place extreme demands on compression and sub-picture handling. The work on boundary padding, sub-bitstream extraction, and tile alignment positions Indian inventions to serve this emerging segment as it matures.

The white space

Even with the current surge of activity, significant opportunity remains. Real-time encoding for live streaming at scale—think millions of concurrent viewers on mobile networks—still needs lighter, smarter codec decisions that can run on edge devices without draining batteries. AI-assisted prediction, where a lightweight model learns content-adaptive modes, is a frontier where Indian innovators could combine their signal-processing depth with machine learning.

Energy-efficient codec implementations for smartphones and IoT cameras are another open field. The work on conditional signaling and simplified candidate lists already points in this direction, but there is room to co-design hardware-friendly algorithms that skip complex searches entirely when the content allows. Sub-picture handling for VR and cloud gaming, where only a viewport is rendered, is also ripe for novel padding and extraction techniques that reduce wasted computation.

Finally, as codec standards evolve toward the next generation, the interplay between transform coding and non-rectangular partitions—like L-shaped blocks—offers a fresh canvas. Indian inventors are already probing transforms adapted to these shapes, and the white space includes integrating such tools into practical, royalty-free codec implementations that can gain wide adoption.

Explore the innovators

The specific inventors, patents, and companies driving these advances in India are mapped in detail on Deeptech Navigator. From novel MPM list designs to adaptive filtering breakthroughs, the platform connects you directly to the people and intellectual property shaping the next wave of video compression. Dive in to see who is building the codec components that will power tomorrow’s streaming experiences.

In our data

Sources

• Video codec ↗
• The overview of video coding technique ↗
• Advanced Video Coding - an overview ↗
• Vibe Coding for Supply Chain Explained ↗
• Digital Supply Chain Management in the Videogames Industry ↗
• Digital Supply Chains ↗
• Advanced Video Coding Market Size & Growth Forecast, 2032 ↗
• Advanced Video Coding Market Size, Competitors & Forecast ↗