İNŞaat mühendiSLİĞİ anabiLİm dali

Orıgınal Mathematıcal Methods For Image Codıng And Compressıon In Modern Communıcatıon Systems
In recent years, an intensive increase has occurred in the formation of image data in parallel with the advances in information technologies. Increasing raw or uncompressed images require high level of source for conduction band width and storage capacity in communication systems. Especially the increase in multimedia applications has brought out the need of more productive methods, high compression ratio and instant communication technology for coding images. With this thesis, a new contribution is offered for the earlier studies carried out for the modeling of images with energy and structure functions. The most important feature of the new method is that it shows better performance when the images obtained as a result of compression are compared with current methods discrete cosinus transform based jpg and discrete wavelet transform based jpg2000.
In this proposed method, the image modelled through vector banks predefined in various block dimensions formed by using training images is recreated. In the proposed method, the image is primarily divided into maximum sized blocks for operation convenience. Afterwards, these blocks are divided into predefined minimum sized blocks which are sub limit depending upon threshold value determining the compression quality by using quad tree algorithm. Each block is modelled as a composition of pattern blocks, energy blocks and scaling constant.
Energy and Pattern blocks have been found for various sized blocks for databases with the method. In this study, many images have been analyzed. As a result of the analysis, a similarity between energy blocks and structure blocks has been observed and it is accepted that these can be eliminated with an appropriate method.  is created from the remaining Energy and Structure blocks after the elimination process. This is placed in the receiver and transmitter found in the bank communication system. In this way, saving in the fields of communication band and storage has been achieved in communication systems considerably. In , there are appropriate energy and structure blocks that can represent the specific image in the best way in terms of the smallest frames. Dependently, a scaling constant can be formed. As a consequence, it has been shown that an image block vector  regarding any image is represented as the multiplication of energy block  and structure block  functions with a  scaling coefficient.
Thanks to this method developed as a result of the thesis, a better compression performance than the current methods has been achieved and contributions have been made to meet the requirements regarding the storage capacity and band width that are more needed with each passing day. It is also thought that this is an important source for minimizing this need thanks to the new algorithms to be developed based on this method in the future.