Weighted Prediction Encoding Feature Offset Calculation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image encoding techniques require traversing all pixels of frames multiple times to determine whether to use weighted prediction, leading to high computation costs and data dependency, making it difficult to implement in pipelined hardware.
Innovation Solution
An image encoding device and method that calculate feature values for consecutive frames and determine offset values to decide whether to activate weighted prediction operation using a simple calculation, reducing computation and data dependency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all pixels of current frame and reference frames are traversed multiple times to generate parameters for decision making, then the accuracy of determining weighted prediction usage is improved, but the computation amount becomes overly large
Solution Approach 1:
The patent extracts only the necessary feature values from frames instead of processing all pixels. Specifically, it extracts feature values from the current frame and reference frames that are sufficient to determine weighted prediction usage, thereby reducing computation while maintaining decision accuracy.
Solution Approach 2:
The patent performs partial action by calculating only the necessary feature values needed for weighted prediction determination rather than fully processing all pixel data. This partial processing approach reduces computational load while still providing sufficient information for accurate decision making.
2Measurement precision
If all pixels of current frame and reference frames are traversed multiple times to generate parameters, then the accuracy of determining weighted prediction usage is improved, but data dependency is resulted during computation
Solution Approach 1:
The patent segments the computation process into independent feature value extraction and parameter calculation stages. By dividing the processing into separate, independent steps that do not require multiple traversals of all pixel data, it reduces data dependency while maintaining measurement precision.
3Measurement precision
If complex parameter calculation is performed to determine weighted prediction usage, then the accuracy is improved, but the image coding technique cannot be easily implemented by pipelined hardware
Solution Approach 1:
The patent segments the complex parameter calculation into simple, independent feature value extractions that can be performed in parallel. This segmentation enables pipelined hardware implementation where different frames can be processed at different stages of the pipeline simultaneously, improving both ease of manufacture and processing efficiency.
Solution Approach 2:
The patent performs preliminary action by pre-calculating and storing feature values from frames before the actual weighted prediction decision is made. This preliminary extraction of feature values simplifies the subsequent decision-making process and enables easier hardware implementation through pre-computed data availability.
Data Source
AI summary
A weighted prediction image encoding method includes: calculating a first feature value of a first frame, a second feature value of a second frame and a third feature value of a third frame; calculating a first offset value according to the first feature value and the second feature value, and calculating a second offset value according to the second feature value and the third feature value; and when both of the first offset value and the second offset value are greater than or less than a predetermined value, encoding the third frame by utilizing weighted prediction operation, wherein the first frame, the second frame and the third frame are multiple frames in chronological order.