研究目的
To propose an automated content selection method for wide color gamut stimuli that objectively characterizes content based on perceptual properties, enabling the selection of representative, diverse, and challenging subsets for various studies.
研究成果
The proposed framework reliably selects wide color gamut source images with diverse color characteristics, validated by high correlation between predicted and actual MOS. The use of cssim as an objective metric and clustering for content selection ensures consistent and unbiased results for QoE studies.
研究不足
The method primarily focuses on color characteristics and may not fully account for other perceptual aspects such as texture or detail loss due to gamut reduction.
1：Experimental Design and Method Selection
The framework involves applying a gamut mapping operator to a wide color gamut reference source image repeatedly to obtain images with successively reduced gamuts. An objective metric measures the perceptual difference score of each gamut-mapped image from the source image. A clustering algorithm is then applied to these scores to obtain content clusters, from which representative samples are selected.
2：Sample Selection and Data Sources
The dataset includes 30 wide color gamut reference images extracted from the HdM-HDR-2014 dataset, converted to low dynamic range RGB color space and gamut-clipped to P3. Additional images from the ArriImageSet and Alexa mini sample footage were used for validation.
3：List of Experimental Equipment and Materials
EIZO ColorEdge monitor for displaying images in a standardized test room.
4：Experimental Procedures and Operational Workflow
Images were displayed side-by-side, cropped to half-width for comparison. Subjects evaluated the color difference between a reference image and its gamut-reduced versions via paired comparison on a 3-point scale.
5：Data Analysis Methods
Pearson correlation coefficient (PCC) was used to compare predicted and actual mean opinion scores (MOS). A sigmoid function was used for fitting objective scores to MOS.
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
