Voicing of animated GIF by data hiding

Abstract

GIF animations are silent image sequences widely used on the web thanks to their wide support and portability. In this work, we propose an original technique based on data hiding, to add sound tracks in GIF animations. Data hiding is usually used to embed security codes in a host medium to prevent from illegal copying or to protect copyrights (watermarking) or to send secret messages to a dedicated receiver (steganography). We propose to use host GIF images as a transmission channel to convey hidden sound bits with lowest perceptual image distortion and without altering the wide portability of the GIF format, by means of data hiding. The inserted bits are neither secret nor intended for security issues. They are intended to be played by an audio player synchronously with the GIF player to add sound to the GIF animation. The embedding process is a low complexity, luminance based steganography algorithm, that slightly modifies the pixels colors of the GIF images to insert the sound bits. The extraction of the inserted audio is completely blind: the audio is directly extracted from the pixels of each cover image. The proposed GIF voicing was tested with different GIF sequences (cartoons and real scenes) and no audio degradation was reported while a slight, most imperceptible, color modification was noticed in case of an important amount of inserted data. The cover images have undergone objective quality criteria and informal subjective evaluation and has proved to be of good quality.