| Paper ID | D-1-3.6 |
| Paper Title |
EVALUATION OF THE ENCODING ACCURACY OF THE PQ BASED HDR CONTENT DELIVERY FORMATS |
| Authors |
Asif Siddiq, Riphah International University, Pakistan; Ishtiaq Rasool Khan, University of Jeddah, Saudi Arabia; Jameel Ahmed, Riphah International University Islamabad, Pakistan |
| Session |
D-1-3: Image/Video Coding |
| Time | Tuesday, 08 December, 17:15 - 19:15 |
| Presentation Time: | Tuesday, 08 December, 18:30 - 18:45 Check your Time Zone |
|
All times are in New Zealand Time (UTC +13) |
| Topic |
Image, Video, and Multimedia (IVM): |
| Abstract |
Recently new formats like HDR10 and Dolby Vision have emerged and became the popular choice for delivery of HDR content. HDR10 uses bit-depth of 10 to quantize the data after applying a non-linear transformation known as perceptual quantization (PQ). It has been known that at this precision of just 10 bits, banding artifacts can be observed in the lower luminance regions. However, in general HDR10 has a good visual quality on HDR displays and is widely recognized for now by the display makers and content developers. In this paper, we evaluate the accuracy of the content encoded in HDR10 format, especially when the dynamic range is high. It is found that the accuracy of HDR10 is low when there are extremely dark and bright regions in the image, and therefore while it is a good choice for HDR display, other formats should be used to preserve the original dynamic range for applications such as machine based analysis and tone-mapping for LDR display. We have also proposed a dual layer codec using companding which resolves the known drawbacks of two-layer backward compatible formats. The proposed format does not require metadata for reconstruction of HDR image and therefore its decoding function has low computational cost. Moreover, its encoding accuracy is higher than, not only HDR10, but also the existing state of the art popular formats like JPEG-XT and HDR-MPEG |