Image and Video Decoding With Distribution Functions for OOR Values
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image and video compression technologies struggle to maintain quality while efficiently encoding and decoding data, especially when machines analyze video for events and objects, as they often compromise on perceptual quality.
Innovation Solution
Implementing distribution functions to handle out-of-range values in bitstreams, using neural networks to encode and decode data, and updating values based on predefined limits to ensure efficient encoding and decoding processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If existing image and video compression technologies are used, then data compression is achieved, but perceptual quality is compromised
Solution Approach 1:
The patent changes the parameter representation by using distribution functions (CDFs) to model the statistical characteristics of compressed data values. Instead of directly encoding raw values, the system encodes parameters of distribution functions that describe the probability distribution of values in the compressed data, thereby preserving perceptual quality while maintaining compression efficiency.
Solution Approach 2:
The patent segments the encoding process into multiple stages: first encoding the distribution function parameters, then handling out-of-range values separately using a second distribution function. This segmentation allows the system to optimize different parts of the data representation independently, preserving quality information that would otherwise be lost in traditional compression.
2Measurement precision
If distribution functions are used to handle out-of-range values, then encoding accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-defining distribution functions and their parameters before the actual encoding process. The system prepares the statistical models in advance, so that during encoding, it only needs to select and apply the pre-computed distribution functions rather than computing them from scratch, thereby reducing real-time computational complexity while maintaining encoding accuracy.
Solution Approach 2:
The patent introduces distribution functions as intermediary elements between the raw compressed data and the final encoded representation. These distribution functions act as mediators that simplify the encoding process by transforming complex value distributions into manageable parameter sets, reducing the computational burden while preserving encoding precision.
3Reliability
If out-of-range values are encoded using multiple distribution functions, then data fidelity is maintained, but encoding time increases
Solution Approach 1:
The patent applies dynamics by making the selection of distribution functions adaptive and dynamic rather than static. The system dynamically selects which distribution function to use based on the characteristics of the input data and the range of values being encoded. This dynamic approach allows the system to maintain high data fidelity by choosing the most appropriate distribution function for each segment of data, while avoiding the unnecessary computational overhead of applying multiple distribution functions to all data uniformly.
Data Source
AI summary
The embodiments relate to method for encoding and decoding. The method for decoding comprises receiving an input bitstream; determining a first distribution function having an input range with a minimum limit value and a maximum limit value; determining a second distribution function having an input range with a minimum limit value and a maximum limit value; decoding a value from the bitstream using a first distribution function; determining when the value equals with the OOR-value of the first distribution function; whereupon the method further comprises defining a positive value variable to be the maximum value of the input range of the first distribution function; defining a negative value variable to be the minimum value of the input range of the first distribution function; decoding the value from the bitstream using the second distribution function, whereupon the method further comprises determining when the decoded value is the OOR value of the second distribution function, in which case the positive value variable is updated with the maximum limit value of the second distribution function; and the negative value variable is updated with the minimum limit value of the second distribution function; a value is decoded from the bitstream using the second distribution function; determining when the decoded value is different from the OOR value and is greater or equal to zero, whereupon the positive value variable is updated with the decoded value; and determining when the decoded value is different from the OOR value and smaller than zero, whereupon the negative value variable is updated with the decoded value; creating a representable media according to the positive and/or negative value variables.