Quality enhancement in image enlargement
Interpolation methods, usually employed in image enlargement, cause a degradation of the image quality in proximity to the edges. In fact, the best interpolation leads to fuzziness effects often due to the image features ignored during the process. In order to have a contained definition loss after the enlargement operation, it is necessary to consider the original image as a subsampled image of the enlarged one. It follows that the image needs to be represented by its low and high frequency components together. A method able to consider and provide the above mentioned image components is the bidimensional Fast Fourier Transform (2D-FFT). In the paper a strategy of partitioning & overlapping is proposed to reduce the computational complexity, conditioned by the number and the accuracy of the operations, when the 2D-FFT is applied to the whole image. So, the image enlargement achieves the goal with low definition loss and without expensivecomputational load. The quality enhancement is evaluated in terms of extracted edges.
- 1.A. V. Oppenheim, R.W. Schafer: Digital Signal Processing. (Prentice Hall International Inc., New Jersey, 1975).Google Scholar
- 2.F. Marino, G. Mastronardi: Techniques of Image Enlargement. Invited paper at the 18th International Conference MIPRO'95, sponsored by IEEE (Croatian Section), Rijeka, Croatia, May 22–25, 1995, MIS 2, pp. 1–13.Google Scholar
- 3.G. Mastronardi, F. Spilotros: Image Expansion Operators based on the Hadamard Pyramidal Coding. Progress in Image Analysis and Processing III, (Edited by S. Impedovo, World Scientific, 1993), pp. 111–114.Google Scholar
- 4.G. Mastronardi, S. Sangiovanni: An Adaptive Filter for Image Processing in OCCAM2. Proc. of the ISMM Int. Symp. on Industrial, Vehicular and Space Applications of Microcomputers, New York, 1990, pp. 69–72.Google Scholar
- 5.M. Sonka, V. Hlavac and R. Boyle: Image Processing, Analysis and Machine Vision. (Chapman & Hall, Cambridge, 1994).Google Scholar