Brightness Variation Model Selection for Video Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing picture coding and decoding methods, such as H.264, face inefficiencies in compressing sequences with non-uniform brightness variations, like local lighting changes or flashes, due to limited flexibility in modeling brightness variations across a picture.
Innovation Solution
A method for decoding and coding picture sequences that selects the most appropriate brightness variation model from a set of models to minimize modeling error, specifically by calculating and comparing parameters for additive and multiplicative models based on neighboring pixel data, allowing for improved prediction and compression in the presence of local brightness variations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single brightness variation model is used for all blocks referring to the same reference picture, then the device complexity is reduced, but the manufacturing precision of brightness compensation deteriorates when local brightness variations occur
Solution Approach 1:
The picture is divided into multiple slices, and each slice can have its own brightness variation model parameters. This segmentation allows different regions of the picture to be compensated with region-specific models, improving local brightness compensation precision while maintaining manageable device complexity through hierarchical organization
Solution Approach 2:
Different brightness variation model parameters are applied to different slices or picture regions. Each region can have optimized parameters (multiplicative and additive) tailored to its specific brightness characteristics, enabling precise local brightness compensation rather than applying a uniform model across the entire picture
2Manufacturing precision
If multiple reference pictures are used to handle local brightness variations, then the brightness compensation precision is improved, but the device complexity increases due to profile and level limitations
Solution Approach 1:
Instead of using multiple reference pictures, the invention changes the parameters (multiplicative and additive values) of the brightness variation model for different slices. This allows the same reference picture to be used with different compensation parameters tailored to each slice's brightness characteristics, achieving high precision without increasing reference picture management complexity
Solution Approach 2:
The solution moves from the dimension of multiple reference pictures (temporal dimension) to the dimension of multiple parameter sets per slice (parameter space dimension). By varying the brightness model parameters across slices rather than using multiple reference pictures, the system achieves flexible local adaptation without the overhead of managing multiple reference pictures
3Adaptability or versatility
If different brightness variation models are assigned to each reference picture, then the adaptability is improved, but the manufacturing precision deteriorates because the model cannot vary for blocks predicted from the same reference picture
Solution Approach 1:
The picture is segmented into slices, and each slice is assigned its own brightness variation model parameters. This enables the model to vary locally across different slices even when they reference the same reference picture, achieving both adaptability and local precision simultaneously
Solution Approach 2:
Each slice can have customized brightness variation parameters (multiplicative and additive) that reflect the local brightness characteristics of that region. This local quality approach ensures that blocks within the same slice receive consistent and appropriate brightness compensation, improving manufacturing precision while maintaining adaptability
Data Source
AI summary
A method for decoding a stream of coded data representative of a sequence of pictures is described. The method comprises the following steps: decoding at least one part of the stream into decoded data, selecting a brightness variation model representative of a brightness variation between the pictures of the sequence from a set of at least two brightness variation models, and reconstructing the picture data from decoded data by taking into account the brightness variation model. A method for coding a picture sequence and a data structure are further disclosed.