With the exploitation of ocean and the need of national defence, visual detection technology as an effectively supporting of underwater non-acoustic detection was received broad attention. In recent years, advances in autonomous underwater vehicle (AUV) and underwater communication technology had promoted a surge of research in the field of underwater video.Effective data storage and data transmission in acoustic or radio channel were the basis of visual detection. Due to limited bandwidth of acoustic channel and storage space for video data, underwater video compression technology became the key to underwater visual detection. Because of the better time-frequency localization and multiresolution analysis, wavelet transform was widely used in image compression.With the need of underwater visual detection, wavelet-based underwater video compression was introduced in this paper. The basic theory of wavelet transform and the characteristics of wavelet image were analyzed at first. Then motion compensated prediction and embedded zerotrees wavelet algorithm were studied. Because of linear phase, biorthogonal wavelet is more appropriate than single-orthogonal wavelet in image compression. Because of the slow motion of AUV, the underwater video had strong correlation. After analysis several motion estimation algorithms, adaptive rood pattern search method was more suitable for underwater video compression. According to the low contrast and low detailed nature of underwater image, embedded zerotrees coding algorithm was improved in this paper. Combined with motion compensated prediction, wavelet-based underwater video compression was designed. According to the experiment of underwater video compression, the algorithm in this paper was met requirements. |