Implementation on FPGA of a Bayer filter for RAW compression in the future JPEG-XS standard
Files
Laurent_37901200_2017.pdf
Closed access - Adobe PDF
- 5.33 MB
Details
- Supervisors
- Faculty
- Degree label
- Abstract
- The digital video keeps improving its quality through the increase of the resolution or the higher frame rate. As a result, the higher required bandwidth complicates the transportation and the storage of the contents. For the past few years, a call for real-time compression has been emitted in order to decrease the bandwidth in the video links. Tico, a low latency and low complexity codec created by IntoPix, can be inserted in the video transport chain in this optic. In the case where the content is transferred starting from the acquisition by the camera, the bandwidth can be further decreased. Indeed, instead of compressing the demosaiced three-components picture, the picture captured by the photosensitive sensors in the Bayer configuration can be processed. This kind of pictures is qualified as RAW. The advantage is the reduction of the quantity of information to compress. The compression in this way required a Bayer filter that creates the different components starting from the image acquired by the sensors. This Master thesis studies the hardware implementation that Tico requires in order to compress such RAW pictures. First, as the data of the RAW content present a higher bit depth than Tico can currently handle, some adaptations have to be realized. Then, the Bayer filter of the first profile is implemented. Finally, the decorrelation of the constructed components is also analyzed in this work. A method based on gradients is proposed to improve the color transform and thus the compression.