Research On Underwater Acoustic Image Compression Based On Perceived Quality Prediction

Image compression is an important subject in underwater image engineering.Different image compression strategies can cause different image degradation features both in storage and transmission.To maximize the quality of image transported in the limited channel resource is a focus problem of common concern.At present,the degradation curves of the conventional image compression exist 'cliff effect',and how to find the best balance between channel resources and image quality is a concern.In view of this problem,this paper uses the perceived quality instead of the PSNR to evaluate the image compression performance.The relationship between the underwater acoustic channel resources and the image compression performance is precisely balanced by establishing a perceptual quality prediction model.We use the image quality measure(IAM)as an image feature,and based on the summaries of compression quality variation regularities of different images,the complex three-dimensional relationship between IAM,BPP and the compression image quality is decomposed into the combinations of several two-dimensional relations by using the quality service vector(QV)method.Through the analysis and fitting of these two-dimensional relations,a compression image quality prediction model is set up.By quickly estimating the image IAM,we can predict the quality of compressed images with the help of quality prediction model.Different compression schemes are needed on indifferent system conditions.Classify the images into different levels of perceived quality based on the analysis and comparison of different coding strategies.Therefore different coding methods are used for different levels of images according to the predicted image quality got form the prediction model to realize the best compression of underwater images.At the same time,the inverse function of the image quality prediction model can be used to provide a desired visual quality in lossy compression of images.Because of the different influence of the compressed bit stream on the reconstructed image quality,the inverse function also be used to classify the compression stream into different g~rades,and the Non-peer protection is adopted to ensure the transmission image quality.