Visual Data Coding With Integer Mask Samples
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Neural network-based image and video coding technologies face challenges in achieving improved coding quality due to susceptibility to coding errors and inefficiencies in processing, particularly when using floating-point numbers in mask samples.
Innovation Solution
The method involves determining mask samples using integer operations on probability parameter samples and values associated with latent representations, ensuring the mask samples are integers, thereby reducing susceptibility to coding errors and improving coding quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If floating-point numbers are used in mask samples, then the mask and scale process can be performed, but coding errors occur and coding quality deteriorates
Solution Approach 1:
The patent changes the data type parameter of mask samples from floating-point to integer. This parameter change eliminates coding errors because integer operations are exact and device-independent, while floating-point operations introduce precision losses and device-dependent results. The mask sample is computed using integer arithmetic operations on integer inputs, ensuring consistent results across different devices.
2Ease of operation
If floating-point operations are used in neural network-based coding, then processing can be performed, but device independence is lost and coding errors increase
Solution Approach 1:
The patent substitutes floating-point arithmetic with integer arithmetic in the mask sample computation. This substitution replaces the imprecise floating-point mechanical system with an exact integer arithmetic system. The integer operations are performed on probability parameter samples and values to produce mask samples that are guaranteed to be integers, eliminating device-dependent precision issues.
3Reliability
If integer operations are used for mask samples, then device independence is achieved and coding errors are reduced, but computational complexity may increase
Solution Approach 1:
The patent changes the operational parameter from floating-point arithmetic to integer arithmetic. This parameter change improves device independence and reliability while the computational complexity remains manageable because the integer operations are performed on probability parameter samples that are already derived from the neural network processing. The integer operations are standard arithmetic operations that do not significantly increase computational burden.
Data Source
AI summary
Embodiments of the present disclosure provide a solution for visual data processing. A method for visual data processing is proposed. The method comprises: determining, for a conversion between visual data and a bitstream of the visual data with a neural network (NN)-based model, a mask sample by performing one or more integer operations on at least one probability parameter sample and at least one value, wherein the mask sample is used in a mask and scale process of the NN-based model, the at least one probability parameter sample is associated with a latent representation of the visual data, and each of the at least one probability parameter sample and the at least one value is an integer; and performing the conversion based on the mask sample.