Codec Performance Curve Fitting for Monotonic Bitrate Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for comparing codec performance metrics, such as Bjontegaard Delta PSNR and Bjontegaard Delta Rate, introduce significant deviations due to non-monotonic data points and limited applicability to non-overlapping intervals, leading to inaccurate and inflexible measurements.
Innovation Solution
A method involving obtaining a monotonic target function with parameters to represent a nonlinear relationship between codec performance metrics and bitrate, allowing for accurate and flexible performance comparisons through logistic regression or other monotonic functions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If interpolation is used to obtain relationship curves from multiple data points, then the curve can be constructed to connect data points, but significant deviations are introduced for non-monotonic data points leading to inaccurate measurement
Solution Approach 1:
The patent transforms the relationship between codec performance metric and bitrate from a direct interpolation approach to a monotonic parameterization approach. By introducing a monotonic parameter t that maps to both the performance metric and bitrate through separate monotonic functions, the method eliminates the need for direct interpolation between data points, thereby avoiding the significant deviations introduced by traditional interpolation while maintaining ease of curve construction.
2Adaptability or versatility
If interpolation is used to compare overlapping parts in relationship curves, then comparison can be performed, but significant randomness is introduced in comparison results
Solution Approach 1:
The patent introduces a monotonic parameter t as an intermediary that mediates between the performance metric and bitrate. Instead of directly comparing overlapping parts of relationship curves through interpolation, the method uses the monotonic parameterization to establish a reliable mapping between different codecs' performance curves, eliminating the randomness in comparison results while maintaining broad adaptability for comparing any interval.
3Device complexity
If traditional interpolation methods are used, then the process is simple, but the method cannot accurately measure performance when data points are non-monotonic
Solution Approach 1:
The patent segments the relationship between performance metric and bitrate into two separate monotonic functions: one mapping the monotonic parameter t to the performance metric, and another mapping t to the bitrate. This segmentation allows each function to be independently optimized for monotonicity, ensuring accurate performance measurement even when data points appear non-monotonic in the traditional sense, while keeping the overall processing complexity manageable through standard numerical optimization techniques.
Data Source
AI summary
This application provide a method for processing a codec performance metric performed by an electronic device. The method includes: obtaining a plurality of pieces of relationship data between a specified codec performance metric and a bitrate; obtaining a monotonic target function set for the specified codec performance metric and the bitrate, the target function including a plurality of parameters for representing a shape of a function curve; determining values of the plurality of parameters based on the plurality of pieces of relationship data and the target function; and determining a nonlinear relationship between the specified codec performance metric and the bitrate based on the target function and the plurality of parameters.