Fractal Based Multiple Description Image Coding

Fractal image coding is a new image coding technique which was proposed in the recent years and it has attracted the attention of many researchers for its novel scheme and its high compression performance.Network congestion and delay sensibility pose great challenges for multimedia communication system design.This creates a need for coding approaches combining high compression efficiency and robustness.Multiple description (MD) coding has emerged as an attractive framework for robust transmission over unreliable channels.In this dissertation,we major in the theory of fraetal image coding with the main purpose of improving its performance further.We also major in multiple description method.Fractal is applied in MD coding in this thesis.The main works involved in the dissertation are:1.Based on the same scale resolution similarity existing in the image,an improved fractal image coding scheme is proposed.A convergence proof of the new iterated function system is given.We have also optimized the coefficients of iterated function system.The proposed scheme obtains 1.5dB improvement compared with traditional fractal image coding.Compared with the scheme proposed by Zhao,almost the same performance can be achieved,but the proposed method does not need the limited condition like Zhao's.So the proposed method should be faster than Zhao's.2.In this thesis,based on the theory of fractal image coding--the collage theory and the fixed point theory,an optimization fractal image coding method using the collage and the reconstruction error is presented.After the optimization scheme,the fractal decoded image quality can be improved 2.5dB so that the Fisher's quadtree scheme achieves better performance than that of JPEG after optimization.3.A transform which exploits the different resolution redundancy is proposed. The transformation matrix is given in this thesis.We also prove that the matrix is ort.hogonal.For easy transformation construction,the concept and theory about H product are extended.Using the extended H product,we can construct the transformation matrix easily.Compared with traditional fractal image coding,the proposed coder based on the new transform can improve the quality of the reconstructed images effficiently(more than 2dB).4.A multiple description method for fractal image coding is proposed in this thesis. The different coefficients of iterated function system are partitioned into descriptions using different methods.Compared with the classical multiple description image coding method,at the same bit rate,the central decoded image quality and side decoded image quality are improved 2.5 dB and 1.5 dB respectively.We also present a multiple description method based on fractal and wavelet.Two fractal coding methods are used to deal with wavelet coefficients of different subbands.When one description is lost,these two fractal coding schemes have the corresponding recovering methods to recover the lost information.We also propose a method to insert redundancy into descriptions according to R-D function.The results show that this scheme can achieve better performance than other multiple description methods.5.A new multiple description scheme based steganography is proposed.The redundancy between descriptions is embedded into the mainly description so that a lot of bits can be saved.Compared with other redundancy appending methods,at the same central decoded image quality,the side decoded image quality can be improved about 3-4 dB.And at the same side decoded image quality,the bit rate of the proposed method can be saved more than 15%.