Film Grain Encoding via Noise Model Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video encoding schemes struggle to efficiently compress video data containing film grain, often resulting in either high bitrate requirements to preserve grain quality or undesirable visual artifacts and temporal instability at reasonable bitrates.
Innovation Solution
A method involving the determination of noise templates and autocovariance values to select an appropriate noise model database entry, allowing for effective characterization and encoding of film grain, which includes calculating first and second autocovariance values and using these to select noise model parameters for improved encoding and decoding processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional video encoding schemes are used to preserve film grain, then grain information is retained, but bitrate requirement increases significantly
Solution Approach 1:
The patent extracts film grain from the video signal by calculating the difference between the original video signal and the decoded signal. This extracted grain is then encoded separately using a dedicated grain encoding process, allowing it to be preserved without requiring high bitrate for the entire video stream. The grain extraction is achieved through residual signal analysis after conventional decoding.
Solution Approach 2:
The patent segments the video encoding process into two independent parts: the main video content encoding and the grain component encoding. By separating the grain from the video signal and encoding them independently with different encoding parameters, the system can preserve grain quality while using efficient bitrate allocation for each component separately.
2Productivity
If conventional video encoding is used at reasonable bitrate, then transmission efficiency is maintained, but grain quality deteriorates with visual artifacts and temporal instability
Solution Approach 1:
The patent employs feedback mechanisms where the decoded signal is fed back into the encoding process to extract grain information. The extracted grain is then re-encoded and added back to the decoded signal during reconstruction. This feedback loop ensures that grain characteristics are captured and preserved throughout the encoding-decoding-encoding process, maintaining grain quality at reasonable bitrates.
Solution Approach 2:
The patent performs preliminary grain extraction and characterization before the final encoding stage. By analyzing the residual signal early in the process and characterizing grain properties (such as autocovariance values), the system can prepare grain-specific encoding parameters in advance, ensuring grain quality is maintained without compromising overall transmission efficiency.
3Device complexity
If film grain is treated as part of video data for compression, then encoding process is simplified, but compression efficiency decreases due to random nature of grain
Solution Approach 1:
The patent changes the encoding parameters specifically for the grain component based on its statistical characteristics. By analyzing grain properties such as autocovariance and adapting encoding parameters (quantization steps, block sizes, prediction modes) to match grain's random nature, the system achieves efficient compression of grain without using the same parameters as conventional video encoding.
Solution Approach 2:
The patent introduces an intermediary grain analysis and characterization module between the conventional video decoder and encoder. This intermediary analyzes the residual signal to extract and characterize grain, then provides grain-specific parameters to the encoder. This mediator enables specialized grain processing without complicating the overall encoding architecture.
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
A method of processing an image is proposed, which comprises: determining, based on the image, one or more noise templates, wherein each of the one or more noise templates comprises noise pixels representing noise contained in the image; calculating one or more first autocovariance values, based on the noise pixels of at least one of the one or more noise templates; based on the one or more first autocovariance values, selecting an entry of a noise model database among database entries which respectively comprise values of noise model parameters corresponding to a noise model.