The Introduction of Supplementary Material

1. The SAJ-TSCI dataset
The semantic aware just noticeable difference based text screen content images (SAJ-TSCI) dataset is provided in the supplementary material to support the research on perceptual distortion in TSCI encoded using versatile video coding (VVC). This dataset includes source images, distorted images, eye tracking maps, and just noticeable difference (JND) data.

2. The source images and distorted images
The SAJ-TSCI dataset contains a total of 24 source images. Source images all contain webpage content, with the spatial resolution range is 320 to 1210 pixels. The color space of source images is YCbCr 4:4:4, and the bit depth is 8, maintaining consistency across the dataset.

The SAJ-TSCI dataset contains a total of 744 distorted images, with each source image corresponding to 31 distorted images with the quantization parameter (QP) being 28 to 58. In order to not identify the authors' identities, we will cover the information that may identify the authors' identities in the source images with a black box and remove it later.The source image is coded using VTM16.2 recommended by the VVC standard, referring to the default settings in the  configuration file. Due to the maximum limit of 50MB for supplementary material only the distorted images corresponding to the No. 10 to No. 22 are provided, and No. 13, 15, 19, 20  images that may  identify the authors' identities have been removed. Referring to the encoding process and source images provided in the Section 3.1 of manuscript and supplementary material, other distorted images can also be easily obtained.

3. The eye tracking map
The eye tracking maps are provided in the SAJ-TSCI dataset. Due to the maximum limit of 50MB for supplementary material, only the eye tracking maps of Subject 1 and Subject 2 are provided for distorted images with QP of 40-55 corresponding to the No. 22 Image. Part of the eye tracking maps of Subject 1 have been analyzed in the Section 4.4 of manuscript. It should be noticed that, the distorted images with QP exceeding 55 cannot provide semantic information in the perception process of subjects, and their observation paths are similar to that of the distorted images with QP=55. And the distorted images with QP below 40 can provide all semantic information in the perception process of subjects, and their observation paths are similar to that of the distorted images with QP=40.

4. The JND data
After clustering the data from T-JND and S-JND experiments, the JND points of each distorted images are obtained. Please refer to the Section 4.1 of manuscript for the specific methods. The T-JND and S-JND points can be used for subsequent researches, have important guiding significance for the subsequent development of efficient JND models suitable for TSCI compressed by VVC.



