LC-ELM-Based Gray Scale Image Watermarking in Wavelet Domain

The applicability of local coupled extreme learning machine (LC-ELM) onto gray scale image watermarking based on discrete wavelet transform (DWT) is described in this work. The learning ability along with generalization toward noisy datasets is examined on synthetic datasets by Y. Qu. (Neural Comput Appl. doi 10.1007/s00521-013-1542-4, 2014). Motivated by the work of (Neural Comput Appl. doi 10.1007/s00521-013-1542-4, 2014), LC-ELM is successfully applied onto image watermarking to test the imperceptibility, and resistance against image processing operations verifies the robustness. Image datasets formed using the selected blocks of approximate subband based on fuzzy entropy are supplied as input to LC-ELM in the training procedure. The binary watermark is embedded into the predicted value obtained using the nonlinear estimation function obtained through trained LC-ELM. The generalization performance of LC-ELM against noisy datasets onto image watermarking is examined by successful extraction of watermark against a number of image operations such as filtering median, average filtering, compression based on JPEG, contrast enhancement, scaling cropping, etc. on different textured images.