Adaptive Video Thinning for Machine Vision Bandwidth Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Videos often consist of large amounts of data, with some portions being less important for machine vision tasks, leading to costly transmission and storage, and inefficient use of bandwidth.
Innovation Solution
A video encoding method that selectively thins videos by deciding which pictures to remove or encode with higher quantization parameters, prioritizing important frames for machine vision tasks, and using rules for reconstructing features in decoded pictures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all pictures in a video are encoded and transmitted, then complete video quality is maintained, but bandwidth consumption and storage requirements increase significantly
Solution Approach 1:
The patent extracts and removes redundant or less important pictures from the video sequence before encoding. By identifying and eliminating duplicate or highly similar frames, the system reduces the total number of pictures that need to be encoded and transmitted, thereby decreasing bandwidth consumption while maintaining essential video content.
Solution Approach 2:
The patent applies different encoding strategies to different portions of the video sequence. Important pictures that contain critical information for machine vision tasks are encoded with higher quality, while less important or redundant pictures are either removed or encoded with lower quality, optimizing the balance between video completeness and bandwidth usage.
2Measurement precision
If all pictures are encoded with high quality, then machine vision task accuracy is maintained, but transmission cost and processing time increase
Solution Approach 1:
The patent removes pictures that do not contribute to machine vision task accuracy, such as duplicate frames or frames with minimal changes. This extraction process reduces the total number of pictures that need to be processed and transmitted, thereby decreasing processing time while maintaining the accuracy needed for machine vision tasks.
Solution Approach 2:
The patent applies selective encoding quality based on the importance of each picture for machine vision tasks. Pictures containing critical features or changes are encoded with high quality to maintain accuracy, while less important pictures use lower encoding quality, reducing overall processing time without sacrificing task performance.
3Adaptability or versatility
If video data is compressed using standard schemes, then compatibility with various devices is achieved, but important machine vision features may be lost
Solution Approach 1:
The patent performs preliminary identification and protection of machine vision features before standard compression is applied. By detecting and marking important features in advance, the system can ensure that these features are preserved through selective encoding or enhanced protection mechanisms, preventing information loss while maintaining device compatibility through standard compression schemes.
Solution Approach 2:
The patent applies different encoding parameters to different regions or pictures based on their importance for machine vision tasks. Pictures or regions containing critical machine vision features are encoded with parameters that preserve feature integrity, while other areas use standard compression, thus balancing device compatibility with feature preservation.
4Reliability
If more video pictures are transmitted, then better video coverage is achieved, but storage requirements and energy consumption increase
Solution Approach 1:
The patent extracts and removes redundant pictures that do not add value to video coverage, such as duplicate frames or frames with minimal content changes. This reduction decreases the total data volume that needs to be stored and transmitted, thereby lowering energy consumption while maintaining adequate video coverage through the remaining essential pictures.
Data Source
AI summary
A method (400) for thinning a video comprising a sequence of pictures. The method includes the deciding whether or not to perform a video thinning process on a picture of the video. The method also includes performing a video thinning process on the picture of the video as a result of deciding to perform a video thinning process. The method also includes deciding whether or not to perform a video thinning process on another picture of the video. The method also includes, after deciding not to perform a video thinning process on the another picture, encoding the another picture to produce an encoded picture. The method further includes adding the encoded picture to a bitstream.