Image compression by means of cellular neural networks

The requirement of high-quality image compression methods is becoming more and more critical with the flourishing of multimedia applications. The development of the cellular computing, e. g. cellular automata (CA), Cellular Neural Networks (CNN), in the past few decades has revealed various feasibility to implement these kinds of computational frameworks for signal processing. Due to their massive parallel nature, CNN have been proven well suitable for image processing. In this thesis, inspired by Dogaru’s work, a wavelet-based image compression method is proposed. The CNN paradigm is implemented in an image compression scheme as far as possible. In this thesis, the nonlinear dynamics and the parallel computing capability of CNN have been investigated. Different CNN-based algorithms of operations involved in the image compression have been developed. The proposed method has proven a comparable performance in both objective quality and the perceptual quality to that of the JPEG 2000 standard for image compression applications where a high compression ratio is required. The results obtained in this thesis show that the application of a CNN-based image compression method can lead to a high-quality while retaining a low system complexity by taking advantage of the CNN paradigm.