Implementation of Bch Coding on Artificial Neural Network-based Color Image Watermarking

This study suggests a novel watermarking technique that uses artificial neural networks (ANN) and BCH (Bose, Chaudhuri and Hocquenghem) coding together to protect intellectual property rights of a color image. BCH error correction coding method is used to improve the performance of watermark extracting. With this composed technique, image is divided into sub-blocks, and a bit-sequence which is used to train both ANN and the watermark is added to the selected sub-blocks. In the watermark embedding process, besides embedding the bit-sequence as is, the watermark is embedded by encoding the watermark into the original image through BCH coding method. ANN is trained by using the features obtained from the selected sub-blocks to which the bit-sequence is embedded. The extraction process is implemented by using the trained ANN and the features obtained from the selected sub-blocks to which the encoded watermark is embedded. After the extraction process, the extracted watermark is obtained by using BCH decoding method. The results of the study are obtained by using the watermark as is and by encoding with BCH coding method. By using BCH encoding method, watermark extraction success is considerably increased, especially on the watermark extraction cases with low success rates. The watermark is extracted considerably successfully from the watermarked image after various image processing attacks as well.