Weighted determination algoritm of boundary pixels

While working with digital noise reduction techniques, which are based on theory of convolution matrix and used convolution operation, it necessary to use algorithms to bypass boundary pixels in the image pixel matrix. The problem exists because convolution itself algorithm have peculiarity, it mean that peculiarity convolution kernel used to each element of pixel matrix. That feature characterize a lot of classes of methods which used idea of convolution matrix. There are a lot of primitive ways to solve it, but none of these ways made a consensus between economical use of resources and filling border pixels with colour coding, which is not so far from colours of corresponding pixels. The object of research is pixel matrix of image. The subject of study is algorithms for filling boundary pixels when superimposing a convolution matrix on a pixel matrix of an image. The main target is creating of effective filled algorithm for border pixels which are close to code colour to relation pixels for used in convolution matrix. Filled border pixels will use to operation convolution for each pixels original image. Algorithm of filled border pixels by step of applied convolution kernel anchors to the pixel, when pixel accessing in convolution algorithm goes beyond the pixel matrix of the original image. Algorithm takes into account the «special» cases of overstepping and allows to do fast calculation to determine the colour code of the missing pixel. The algorithm is simple to program and easily integrates with the basic convolution matrix algorithm in digital image defects.

1. M. V. Gashnikov, N. I. Glumov, N. Yu. Il’yasova, V. V. Myasnikov, et al., Computer Image Processing Methods, Ed. by V.A. Soifer, Fizmatlit, Moscow, 2003. – 784 p.

2. Yane B. Cifrovaya obrabotka izobrazhenij. Moskva: Tekhnosfera, 2007. – 584p.

3. Fisenko V. T., Fisenko T. U. Komp’yuternaya obrabotka i raspoznavanie izobrazhenij: ucheb. posobie. – SPb: SPbGU ITMO, 2008. –192 p.

4. Forsyth. D, Ponce. J. Computer vision: a modern approach. Upper Saddle River, N.J., Prentice Hall, 2003.

5. Shapiro L. Kompyuternoezrenie / L Shapiro, Dzh. Stokman; Per. s ang. – M.: BINOM. Laboratoriy aznanij, 2006. – 752 p.

6. Bailey D. G. Image border management for FPGA based filters. In: 6th IEEE international symposium on electronic design, test and applications, Queenstown, 17–19 Jan 2011, pp 144–149.

7. Ronald N. B. The Hartley Transform. //Oxford University Press, Inc. 1986 Madison Ave. New York, NY United States.

8. Hirshman I. I., Uidder D. V. Convolution / Hirshman I. I, Uidder D. V. Conversion type svertki, 1958. – 312 p.

9. Oppenheim A., Schafer R. Digital signal processing, 1979. – 283 p.

10. Zaerko, D. V. Weighted method solving of boundary pixels problem in digital noise convolution filtering algorithm / D. V. Zaerko, V.A. Lipnitski // Kodirovanie i cifrovaya obrabotka signalov infokommunikaciyah: materials of the international scientific conference, BSUIR, Minsk, Rep. Belarus, 24 April 2020 г. – P. 59–63.