Online Per-Title Encoding for Low-Latency Live Streaming
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current live streaming technologies face high latency due to the computational expense of determining optimal bitrate-resolution pairs, leading to suboptimal video quality and viewer experience, as per-title encoding is not typically deployed in live streaming applications.
Innovation Solution
The implementation of online per-title encoding using discrete cosine transform (DCT)-based energy functions to extract low-complexity spatial and temporal features, predicting optimal bitrate-resolution pairs for each video segment, and applying these to achieve low-latency encoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If per-title encoding is used to find optimal bitrate-resolution pairs, then video quality is improved, but latency increases due to computational expense
Solution Approach 1:
The patent pre-calculates and stores convex hull data structures for various video complexity levels before live streaming. During live streaming, the system only needs to classify the incoming video's complexity level and retrieve the pre-computed optimal bitrate-resolution pairs, avoiding real-time convex hull computation while maintaining per-title encoding quality
Solution Approach 2:
The patent segments the continuous video stream into discrete complexity classes (e.g., low, medium, high complexity). Each segment type has pre-computed encoding parameters stored in lookup tables. This segmentation transforms the continuous optimization problem into a discrete classification and retrieval process, significantly reducing computational latency
2Manufacturing precision
If per-title encoding is deployed for each video content, then optimal bitrate ladder is achieved, but computational complexity increases
Solution Approach 1:
The system pre-computes convex hulls for representative video samples across different complexity levels and stores them in lookup tables. During live streaming, instead of performing full convex hull computation for each video, the system classifies the video complexity and retrieves pre-computed parameters, reducing computational complexity from O(n log n) to O(1) lookup operations
Solution Approach 2:
The patent changes the parameter space by pre-computing and storing optimal bitrate-resolution pairs for different video complexity levels. This transforms the real-time optimization problem into a parameter retrieval problem, where the system only needs to determine video complexity level and select corresponding pre-optimized parameters
3Loss of time
If fixed resolution-bitrate ladder is used for live streaming, then latency is reduced, but video quality deteriorates
Solution Approach 1:
The patent pre-computes and stores optimal bitrate-resolution pairs for multiple video complexity levels before live streaming begins. During streaming, the system classifies incoming video complexity and retrieves the corresponding pre-optimized parameters from lookup tables, achieving both low latency (via pre-computation) and high quality (via per-title optimization)
Solution Approach 2:
The patent applies different bitrate-resolution parameters tailored to each video's specific complexity characteristics. Instead of using a universal fixed ladder, the system selects from multiple pre-computed ladders that are locally optimized for different video types, ensuring each video receives the most appropriate encoding parameters for its specific content characteristics
Data Source
AI summary
The technology described herein relates to online per-title encoding. A method for online per-title encoding includes receiving a video input, generating segments of the video input, extracting a spatial feature and a temporal feature, predicting bitrate-resolution pairs based on the spatial feature and the temporal feature, using a discrete cosine transform (DCT)-based energy function, and per-title encoding segments of the video input for the predicted bitrate-resolution pairs. A system for online per-title encoding may include memory for storing a set of bitrates, a set of resolutions, and a machine learning module configured to predict bitrate resolution pairs based on low-complexity spatial and temporal features.