1：Experimental Design and Method Selection:

The method involves representing HDR images as third-order tensors, applying Tucker3 decomposition to generate three feature maps, extracting multi-scale manifold structure features from the first feature map using PCA and OLPP algorithms, and extracting multi-scale contrast features from the second and third feature maps using standard deviation calculations. Support vector regression is used to aggregate features into a quality score.

2：Sample Selection and Data Sources:

Two public HDR image databases are used: the Nantes database with .hdr format images and the EPFL database with .pfm format images. Training sets include 10 non-distorted HDR images from DML-HDR set for Nantes and randomly selected 10 from EPFL for EPFL database testing.

3：List of Experimental Equipment and Materials:

No specific equipment or materials are mentioned; the method is computational and uses software algorithms.

4：Experimental Procedures and Operational Workflow:

Steps include tensor decomposition of HDR images, feature extraction from feature maps at multiple scales, and SVR-based quality prediction. Performance is evaluated using PLCC, SROCC, and RMSE metrics after nonlinear fitting.

5：Data Analysis Methods:

Performance is assessed using Pearson linear correlation coefficient (PLCC), Spearman rank order correlation coefficient (SROCC), and root-mean-square error (RMSE). Data is fitted with a five-parameter logistic function to predict MOS.