| Abstract: |
7 Conclusion And Future Work7.1 ConclusionIn this chapter, we presented a fractal image compression based on a self-organizing map neural network. The algorithm uses the neural networks self organizing map for the classification of the RAW blocks after the transformation to zero-mean and transferring to the unit sphere. The algorithm considers the classification on the eight isometric rotations for the sake of better image quality. The eight isometric processes are done over range blocks rather than domain pools for the sake of a compression speed. The trained neural network using the training set taken from a proper standard – aerial image is proved to work effectively for many other images. The algorithm shows remarkably reduction in compression time without a remarkable reduces the image fidelity. Fractal neural processor is a new block diagram designed to reduce the compression time by implementing the algorithm in a hard ware circuit elements. We discuss the block diagram design of the fractal neural processor and explain the functions of the modules describing the fractal neural processor operations. We present also a new block diagram design called a fractal neural processor, which has the capability to make a significant decreasing in time complexity. We took an example to realize the time difference between a proposed algorithm and a fractal neural processor, which is the rotator to be implemented as a hard ware logic circuits using simulation techniques. The implementation of the fractal neural processor will take place in a digital image processing application i.e. fractal neural processor can be used in aircraft camera (used a real time compression during the using of camera).7.2 Future Work7.2.1 Fractal neural processor implementation and realizationOur dissertation is introducing a schematic diagram for fractal neural processor. The future work is to prove the realization of the fractal neural processor in real work.7.2.2 Fractal Lossless fractal compressionFractal image compression is classified as a lossy compression. The future work is to try to be lossless compression.7.2.3 Fractal Video Compression Using Neural NetworkOur algorithm is dealing with still- images. The future work is to manipulate the image sequence image which is called video images.7.2.4 Extracting the algorithm to work in color imagesOur algorithm is working on grey scale images. The future work is to extend the work to capture the color images.7.2.5 Using linear / non linear transformationThe fractal idea is depending on affine transformation meaning that C= Ax + B. the future work is convert the relation into a nonlinear relation.7.2.6 Using flexible block to block transformation rather than a twice size relationIn the fractal image compression mapping from domain blocks to range blokes is the domain blocks twice in size compared to the range blocks. The future work is to try any size between the domain blocks and range blocks7.2.7 Block classification is not necessarily to be square, it may be rectangleBlocking technique in the fractal image compression is to divide into a set of square blocks. Future work is to divide the image into a set of any regular shape.REFERENCES1. A.E.Jacquin, (2002) ”Fractal Image Coding” : A Review, Proceedings of the IEEE, Vol.81, No.10.2. Lisa K. Wells and Jeffrey Travis, (2003) ”LabVIEW for Everyone, Graphical Programming made even easier,”Prentice Hall3. G.Denelli (1999) ”Image Data Compresion”, in V Gappellini and R Macroni, editors, Adanced in image processing and pattern recoginition, page 847-858 SPIE,, Bellingham,wa4. J. Gomes and L. Velho. (1999) “Image Processing for Computer Graphics”. Springer- Verlag.5. H.Kobayashi and L.R.Bahi. (1994) ”Image Data Compression” IBM J.Res.Devl8,No.2 p164-170.6. S. E. Umbaugh, (2002) ”Computer Vision and Image Processing”. Prentice-Hall.7. A. E. Jacquin, (2002, Jan) “Image coding based on a fractal theory of iterated contractive image transformations,” IEEE Transactions on Image Processing vol. 1, pp. 18-30.
|
|
|