US Patent Application 18008789. END-TO-END WATERMARKING SYSTEM simplified abstract
Contents
END-TO-END WATERMARKING SYSTEM
Organization Name
Inventor(s)
Xiyang Luo of Mountain View CA (US)
Feng Yang of Sunnyvale CA (US)
Elnaz Barshan Tashnizi of Toronto (CA)
Ryan Matthew Haggarty of Kitchener (CA)
Michael Gene Goebel of Santa Barbara CA (US)
END-TO-END WATERMARKING SYSTEM - A simplified explanation of the abstract
This abstract first appeared for US patent application 18008789 titled 'END-TO-END WATERMARKING SYSTEM
Simplified Explanation
This patent application describes a method for training an encoder and decoder to generate and decode watermarks in data items. The training process involves generating multiple watermarks for training images, adding distortions to the watermarked images, and adjusting the training parameters based on the error values.
- The patent application focuses on jointly training an encoder and decoder for watermark generation and decoding in data items.
- The training process involves obtaining a set of training images and data items.
- For each training image, a first watermark is generated using an encoder, and then a second watermark is generated by tiling multiple first watermarks.
- The second watermark is used to watermark the training image, and a first error value is calculated.
- Distortions are added to the watermarked image, and a distortion detector predicts these distortions.
- The distorted image is modified based on the predicted distortions.
- The modified image is decoded by the decoder to generate a predicted data item and a second error value.
- The training parameters of the encoder and decoder are adjusted based on the first and second error values.
Original Abstract Submitted
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for jointly training an encoder that generates a watermark and a decoder that decodes a data item encoded within the watermark. The training comprises obtaining a plurality of training images and data items. For each training image, a first watermark is generated using an encoder and a subsequent second watermark is generated by tiling two or more first watermarks. The training image is watermarked using the second watermark to generate a first error value and distortions are added to the watermarked image. A distortion detector predicts the distortions based on which the distorted image is modified. The modified image is decoded by the decoder to generate a predicted data item and a second error value. The training parameters of the encoder and decoder are adjusted based on the first and the second error value.
