Lossless image compression for aerospace non-destructive testing apllications

Abstract

This thesis studies areas of image compression and relevant image processmg techniques with the application to Non-destructive Testing (NDT) images of aircraft components. The research project includes investigation of current data compression techniques and design of efficient compression methods for NDT images.
Literature review was done initially to investigate the fundamental principles of data compression and existing methods of lossless and lossy image compression techniques. Such investigation provides not only the theoretical background, but also the comparative benchmarks for the research project. Chapter 2 provides general knowledge of image compression. The basic predictive coding strategy is introduced at the beginning of chapter 3. Fundamental theories of the Integer Wavelet Transform (IWT) can be found in chapter 4.
The research projects proposed mainly three innovative methods for lossless compression of NDT images. Namely, the region-based method that employs region­oriented adaptation; the texture-based method that employs a mixed model for the prediction of image regions with strong texture patterns; and a hybrid method that utilizes advantages from both predictive coding and IWT coding.
The main philosophy of lossless image compression is to de-correlate the original image data as much as possible by mapping from spatial domain to spatial domain in the predictive coding strategy or from spatial domain to transform domain in the IWT coding strategy. The proposed region-based method aims to achieve the best mapping by adapting the de-correlation to the statistical properties of decomposed regions using the component's CAD model. With the aid of component CAD models to divide the NDT images of aircraft components into different regions based on the material structures, the design of the predictors and the choice of the IWT are optimised according to the specific image features contained in each region having the same material structure. The texture-based method achieves the best de-correlation by using a mixed data model in the region possessing strong texture patterns.
A hybrid scheme for lossless compression of the NDT images of aircraft components is presented. The method combines the predictive coding and the IWT. After region-based predictive coding, the IWT is applied to the error images produced for each decomposed region to achieve further image de-correlation by preserving the information contained in the error images with fewer transform coefficients. The main advantages of using the IWT are its multi-resolution nature and lossless property with integer grey level values in images mapped to integer wavelet coefficients.
The proposed methods are shown to offer a significantly higher compression ratio than other compression methods. The high compression efficiency is seen to be achieved by not only a combination of the predictive coding and the IWT, but also optimisation in the design of the predictor and the choice of the transform according to the specific image features contained in each region having similar material structures.