Research On Acoustic Image Processing Techniques Based On Fuzzy Theory And Spatial Correlation Filtering

Acoustic image plays an important role in human's daily life. People obtainuseful information through analyzing and understanding different kinds of acousticimages, which require high quality and good visual effects of acoustic imagesthemselves. But in many cases, images are often unable to meet the demand, whichappears in low contrast or heavy noise corrupted in images. Most synthetic aperturesonar images have low contrast and high reverberation due to different interferences,as well as imaging and data acquisition system. Therefore it's necessary to find asuitable image enhancement algorithm, to improve image contrast and reduce noise.Acoustic image filtering and enhancement is an important process for advancedimage analysis. Through the preprocessing, image can get better visual effect andmaintain more prominent features. Currently there are many algorithms used in imageenhancement and filtering. This thesis proposes a new image processing technique ofmodified spatial correlation filtering and fuzzy-based enhancement in wavelet domain.Spatial correlation filtering is designed based on the fact that true signal and noiseshave different statistical properties in different scales or with the same scale. On thebasis of traditional spatial correlation filtering algorithm, this thesis presents animproved spatial correlation filtering algorithm using adaptive fuzzy weights medianfiltering. By introducing a local correlation coefficients matrix, this algorithm cangive more accurate judgment between noise and useful signal. At the same time, fuzzyenhancement is carried out on the approximate coefficient in wavelet analysis.Quantitative and qualitative analysis by using the standard Lena image show thatour proposed modified spatial correlation filtering and fuzzy-based enhancementalgorithm improves the image contrast and suppress noises. At last, the algorithm isused to deal with synthetic aperture sonar images, and the results further demonstratethe superiority of the algorithm.