eduzhai > Applied Sciences > Engineering >

GIFnets Differentiable GIF Encoding Framework

  • Save

... pages left unread,continue reading

Document pages: 9 pages

Abstract: Graphics Interchange Format (GIF) is a widely used image file format. Due tothe limited number of palette colors, GIF encoding often introduces colorbanding artifacts. Traditionally, dithering is applied to reduce color banding,but introducing dotted-pattern artifacts. To reduce artifacts and provide abetter and more efficient GIF encoding, we introduce a differentiable GIFencoding pipeline, which includes three novel neural networks: PaletteNet,DitherNet, and BandingNet. Each of these three networks provides an importantfunctionality within the GIF encoding pipeline. PaletteNet predicts anear-optimal color palette given an input image. DitherNet manipulates theinput image to reduce color banding artifacts and provides an alternative totraditional dithering. Finally, BandingNet is designed to detect color banding,and provides a new perceptual loss specifically for GIF images. As far as weknow, this is the first fully differentiable GIF encoding pipeline based ondeep neural networks and compatible with existing GIF decoders. User studyshows that our algorithm is better than Floyd-Steinberg based GIF encoding.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×