研究目的
To develop an adaptive color noise reduction method for non-constant luminance signals in high dynamic range video services to improve compatibility with legacy devices and reduce pre-processing complexity.
研究成果
The proposed adaptive luma adjustment method effectively reduces color noise in HDR video by leveraging CL signal information and color saturation, while significantly decreasing pre-processing complexity by more than two-fold. It maintains compatibility with legacy devices without requiring additional metadata, and can be adapted for future WCG content.
研究不足
The method is specific to NCL signals and may not fully address all types of noise or be applicable to other color gamuts without adaptation. The objective metrics may not perfectly capture visual quality, and the complexity reduction is dependent on the specific conditions set for pixel selection.
1：Experimental Design and Method Selection:
The study involves designing an adaptive luma adjustment method based on the relationship between NCL and CL signals and color saturation levels. It uses theoretical models from MPEG standards and HDRTools software for implementation.
2：Sample Selection and Data Sources:
Test sequences (Market, FireEater, Tibul) with BT.709 color gamut provided by Technicolor are used, with resolutions of 1920x1080 and varying frame rates.
3：List of Experimental Equipment and Materials:
A Dell UltraSharp U3014 monitor for visual observation, Intel Core i7-4790K CPU @ 4.00 GHz with 32 GB RAM for objective testing, and HDRTools 1.0 software for implementation.
4：00 GHz with 32 GB RAM for objective testing, and HDRTools 0 software for implementation.
Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: The method applies luma adjustment only to pixels where the absolute difference between NCL and CL luma values exceeds 0.1 and the minimum RGB value is below 0.0002. It involves transfer functions, color space conversions, quantization, and resampling as per MPEG CfE chain.
5：1 and the minimum RGB value is below It involves transfer functions, color space conversions, quantization, and resampling as per MPEG CfE chain.
Data Analysis Methods:
5. Data Analysis Methods: Objective metrics include mPSNR Y, mPSNR U, mPSNR V, mPSNR, tPSNR X, tPSNR Y, tPSNR Z, tPSNR XYZ, deltaE PSNR, and pre-processing speed in fps. Subjective visual quality assessment is conducted on a monitor.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
