Academia.eduAcademia.edu

Figure 6 – uploaded by Antonio Pinheiro

See full PDFdownloadDownload figure
Figure 9: BD-Rate versus decoding time over xvc Figure 8: BD-Rate versus encoding time over x265

Figure 9 BD-Rate versus decoding time over xvc Figure 8: BD-Rate versus encoding time over x265

Related Figures (8)

Figure 1: Light field images can be represented by the different views captured from one scene.
Figure 1: Light field images can be represented by the different views captured from one scene.
Figure 2: Workflow for pseudo-sequence based light field image compression methods. A) Views are converted to a pseudo-video based on a predefined scan order. B) Pseudo-video is encoded. C) Pseudo-video is decoded. D) Decoded Pseudo-video is ordered to the views based on initial scan order.
Figure 2: Workflow for pseudo-sequence based light field image compression methods. A) Views are converted to a pseudo-video based on a predefined scan order. B) Pseudo-video is encoded. C) Pseudo-video is decoded. D) Decoded Pseudo-video is ordered to the views based on initial scan order.
Implementation setup and Dataset based on MPEG standards like HEVC, but including some extra tools [18]. Four challenging light field images, Greek, Bicycle, Herbs, and Table from 4D Light Field Dataset [19] were selected to be encoded with the selected codecs based on expert review. They are 9 x 9 views with 512 x 512 pixels resolution. The central view of the selected images are shown in Figure 3. The light field image views are firstly converted to 8-bit YCbCr color format and then are downsampled to 4:2:0 chroma subsampling. The views are ordered using a serpentine scan to define a pseudo-video. The second version of xve> is used for the simulations. To set GOP to 16 like other selected codecs, -sub-gop-length is set to 16. To satisfy IntraPeriod= -/ like other codecs, -max-keypic- distance is set to more than number of total frames (default 640 was used).
Implementation setup and Dataset based on MPEG standards like HEVC, but including some extra tools [18]. Four challenging light field images, Greek, Bicycle, Herbs, and Table from 4D Light Field Dataset [19] were selected to be encoded with the selected codecs based on expert review. They are 9 x 9 views with 512 x 512 pixels resolution. The central view of the selected images are shown in Figure 3. The light field image views are firstly converted to 8-bit YCbCr color format and then are downsampled to 4:2:0 chroma subsampling. The views are ordered using a serpentine scan to define a pseudo-video. The second version of xve> is used for the simulations. To set GOP to 16 like other selected codecs, -sub-gop-length is set to 16. To satisfy IntraPeriod= -/ like other codecs, -max-keypic- distance is set to more than number of total frames (default 640 was used).
Table 4: BD-BR metric using PSNR.
Table 4: BD-BR metric using PSNR.
Table 5: BD-PSNR metric using PSNR.
Table 5: BD-PSNR metric using PSNR.
Related topics:
Computer ScienceArtificial IntelligenceComputer VisionData CompressionImage compressionElectronic Imaging
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved