Video Encoding via Segmented Machine Learning Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding approaches do not provide truly optimized encoding for specific video titles, leading to suboptimal performance and quality, especially in adaptive streaming where bandwidth fluctuations require dynamic adjustments.
Innovation Solution
A computer-implemented method and system that analyzes input video files to extract parameters, determines optimized encoding settings using machine learning, and divides the video into time-based chunks for parallel processing, allowing for custom encoder settings that ensure smooth transitions and improved quality metrics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single complexity bucket approach is used for encoding, then encoding speed is maintained, but encoding optimization for specific video titles is lost
Solution Approach 1:
The patent segments the video file into multiple time-based chunks and analyzes each chunk separately to determine its complexity characteristics. This allows the encoding process to be customized for each segment rather than applying a single complexity bucket to the entire video, thereby achieving both speed (through automated chunk-based processing) and precision (through localized optimization).
Solution Approach 2:
The patent applies local quality by determining encoding parameters specific to each time-based chunk based on its individual characteristics. Each chunk receives customized encoding settings tailored to its complexity, motion content, and other local features, rather than a uniform approach across the entire video file.
2Manufacturing precision
If in-depth analysis of the entire video is performed before encoding, then encoding parameters can be optimized, but encoding time increases
Solution Approach 1:
The patent performs in-depth analysis on segmented time-based chunks rather than the entire video file at once. This segmentation allows the analysis and encoding process to be distributed and parallelized, maintaining comprehensive optimization while reducing the time burden through modular processing.
Solution Approach 2:
The patent performs preliminary analysis on each time-based chunk to extract characteristics and determine encoding parameters before actual encoding begins. This preliminary action on small chunks is more efficient than analyzing the entire file, as it allows for targeted analysis of only the relevant features needed for each segment's optimization.
3Ease of manufacture
If coarse customization based on complexity buckets is used, then encoding process is simplified, but truly optimized encoding for specific video titles is not achieved
Solution Approach 1:
The patent maintains ease of manufacture by automating the complexity analysis and encoding parameter determination process for each chunk. The system automatically segments the video, analyzes each segment's characteristics, and applies appropriate encoding settings without manual intervention, thus keeping the process simple while achieving fine-grained optimization.
Solution Approach 2:
The patent achieves video-title-specific optimization by dynamically changing encoding parameters based on the actual characteristics of each time-based chunk. The system adjusts bitrate, resolution, and other encoding parameters according to the measured complexity and content features of each segment, moving beyond fixed complexity buckets to adaptive parameter selection.
4Adaptability or versatility
If multiple encodings are performed for different quality levels, then adaptive streaming capability is improved, but processing complexity increases
Solution Approach 1:
The patent segments the video into time-based chunks and performs encoding for multiple quality levels on these segments. This segmentation allows parallel processing of different quality versions and enables adaptive streaming capability while managing processing complexity through modular, reusable encoding pipelines for each chunk.
Solution Approach 2:
The patent creates a universal encoding framework that can generate multiple quality levels from the same set of time-based chunks. The same chunk segmentation and analysis process serves multiple encoding targets (different bitrates and resolutions), reducing overall processing complexity compared to treating each quality level as a separate full-video encoding task.
Data Source
AI summary
An original input video file is encoded using a machine learning approach. The encoder performs a detailed video analysis and selection of encoding parameters that using a machine learning algorithm improves over time. The encoding process is done using a multi-pass approach. During a first pass, the entire video file is scanned to extract video property information that does not require in-depth analyses. The extracted data is then entered into an encoding engine, which uses artificial intelligence to produce optimized encoder settings. The video file is into a set of time-based chunks and, in a second pass, the encoding parameters for each chunk are set and distributed to encoding nodes for parallel processing. These encoder instances probe-encode each chunk determine the level of complexity for the chunk and to derive chunk-specific encoding parameters. Following completion of the second pass, the results of both passes are then merged to obtain the necessary information for the encoder to achieve the best possible result.