Abstract: This thesis researches the watermarking techniques. Through combining human visual system, an improved algorithm based on DCT transform is proposed. The way is to use the adaptive quantization on the basis of JPEG quantization table, and meanwhile selecting the place of watermark embedding improved. Experimental results show that the selection of embedded place makes algorithm adapted to the image with different general characters and local characteristics, and therefore it generates a lot of mobility. At the same time, good stability and visibility are testified from the experimental data and macroscopic observation.

Key words: human visual system, DCT, quantization