A Self Signature Embedding for Image Security by Using Similarity Based Superior Self Organizing Maps Neural Network

This paper uses self extracted feature value as the signature to embed in the image data. So no additional host image is needed for transaction. Authentic image can be send directly with its self signature. Self sign is mined from the authentic image by using SBS-SOM neural network; hence the sign is unique for each image. No two signs can be same for any different images. Mining of same sign is only possible with the same image. Minute significant values are mined here after the complete training process of the SBS-SOM. The projected system assures imperceptible watermarking than the existing techniques. The authentic image arrival is confirmed at the destination side with watermark extraction and removal from the watermarked image at the other side. Self attested image transformation with the proposed system is one more novel achievement in the issue of secured image transaction.