Research On Efficient Underwater Video Compresion Algorithm

People have increasingly recognized that the ocean, which occupies nearly 75 percent of the earth surface, as a foundation of the life of human beings in the future with the development of ocean exploration. Meanwhile, the ocean considered as the inner space of the earth, plays an important role in military affairs. However, people have limited knowledge about the deep ocean due to the complexity of ocean environment. In recent years, computer vision has increasingly been used as a sensing modality for underwater exploration. Remotely operated vehicles (ROVs) and autonomous underwater vehicles (AUVs) are generally equipped with underwater cameras. ROV transmits video through tether cable, but the tether limits the practicability and maneuverability of the ROVs. AUV is more flexible and can explore in a much larger range. However, it needs to transmit massive video information through acoustic channels. The main obstacle in underwater video transmission is the limitation of low bandwidth of underwater acoustic channel. Current video coding standards cannot be applied to the acoustic channel directly. Taking account of the characteristics of underwater video sequences while maintaining the acceptable perceptual quality of the reconstructed images, the following research work on underwater video coding has been carried out.(1)Underwater images possess the characteristics of lower contrast, less texture details and less color information due to the artificial lightening condition, so a wavelet transform based preprocessing step is proposed. The proposed preprocessing step can reduce the perceptual redundancies in underwater images and the processing time in the subsequent coding procedures. An efficient intra-frame algorithm is implemented using wavelet tree based wavelet different reduction (WDR) after the analyses and comparisons of current intra-frame coding algorithms. The low-pass subband concentrating the most important information of an image is coded by using lossless DPCM based entropy-coding algorithm. We encode the other high frequency subbands using the wavelet tree based WDR method. Experiments show that the proposed preprocessing algorithm and intra-frame coding algorithm are feasible and can improve the compressing rate while maintaining perceptual quality of images.(2)Based on the scenarios in our work, we develop a fast and robust global motion estimation method to overcome the problems in the existing GME algorithms, that is, complicated computation and time-assuming. Our method estimates the final model parameters of image motion, projective model parameters, by combining the result of feature point matching and the refining of rigid model parameters. It can eliminate the incorrect correspondences between images and reduce the computing time for estimating global motion parameters. Numerous experiments illustrate the validity of our method in improving the computing speed.(3)We also propose a hybrid global-local motion compensation scheme to eliminate the temporal redundancy between frames. The predefining of local motion compensation is realized by an assessment on the extracted feature points, thus avoiding the repeated computation of global motion estimation. Therefore, we can reduce the redundancy of the hybrid coding algorithm and computing time. The final information to be encoded and transmitted only includes parameters of global motion model and data of local compensation blocks. The experimental result shows that while maintaining the acceptable perceptual quality of reconstructed images, the proposed hybrid-coding algorithm can provide high coding efficiency, especially suitable for sea floor video compression and transmission at very low bit rates to meet the limited bandwidth of underwater acoustic channels.