Increasing the speed of wavelet image processing based on the Winograd method taking into account decimation

Purpose of research. Wavelet transform is widely used to solve a wide range of digital image processing problems in various applied and scientific and technical fields. At the same time, modern visual information processing systems face the problem of insufficient performance against the background of a rapid increase in digital data volumes. This circumstance requires the development of computationally efficient wavelet processing algorithms suitable for implementation in modern computing devices. This study is aimed at reducing the computational complexity of wavelet image processing based on the use of a modification of the Winograd method. The article proposes the use of a new approach to organizing calculations for one-dimensional filtering with decimation.

Methods. The study used a method for organizing calculations based on the Winograd transform and hardware simulation on a programmable valve matrix in Xilinx Vivado 2018.2 environment using Verilog language for Virtex 7 family model “xc7vx485tffg1157-1”, using standard synthesis and implementation parameters: “Vivado Synthesis Defaults” and “Vivado Implementation Defaults”, respectively.

Results. Experimental modeling of the wavelet transform has demonstrated that the application of the Winograd method in wavelet image processing tasks allows for a reduction the computational delay by 34-63% compared to the direct method when using fourth-order wavelets and by 39-66% when using sixth-order wavelets.

Conclusion. The application of the Winograd method provides a significant increase in the computation speed with some increase in hardware complexity and energy consumption. The results of the study can find wide application in modern signal, image and video processing systems, as well as in the development of machine learning systems.

1. Wu Y., Gao G., Cui C. Improved wavelet denoising by non-convex sparse regularization under double wavelet domains. IEEE Access. 2019; (7): 30659-30671. DOI: 10.1109/ACCESS.2019.2903125.

2. Qin Q., Dou J., Tu Z. Deep ResNet Based Remote Sensing Image Super-Resolution Reconstruction in Discrete Wavelet Domain. Pattern Recognition and Image Analysis. Advances in Mathematical Theory and Applications. 2020; 30(3): 541-550. DOI: 10.1134/S1054661820030232.

3. Soulard R., Carrе P. Elliptical monogenic wavelets for the analysis and processing of color images. IEEE transactions on signal processing. 2015; 64(6): 1535-1549. DOI: 10.1109/TSP.2015.2505664

4. Chen Y., Li D., Zhang J. Q. Complementary color wavelet: A novel tool for the color image/video analysis and processing. IEEE Transactions on Circuits and Systems for Video Technology. 2017; 29(1): 12-27. DOI: 10.1109/TCSVT.2017.2776239.

5. Taranenko Yu. K. Methods of discrete wavelet filtering of measuring signals: algorithm for choosing a method. Izmeritel'naya tekhnika = Measuring equipment. 2021; (10): 14-20. (In Russ.). DOI: 10.32446/0368-1025it.2021-10-14-20.

6. Alimagadov K. A., Umnyashkin S. V. Data augmentation based on wavelet filtering in training neural networks. In: GrafiKon-2023. Trudy Mezhdunarodnoi konferentsii po komp'yuternoi grafike i zreniyu = GraphiCon-2023. Proceedings of the International Conference on Computer Graphics and Vision. Moscow; 2023. P. 437-442. (In Russ.).

7. Simonov E. N., Vinogradov K. M. Image reconstruction by the back projection method using wavelet filtering of projection data in X-ray computed tomography. Vestnik YuzhnoUral'skogo gosudarstvennogo universiteta. Seriya: Vychislitel'naya matematika i informatika = Bulletin of the South Ural State University. Series: Computational Mathematics and Informatics. 2024; 13(2): 5-22. (In Russ.). DOI: 10.14529/cmse240201.

8. Rossinelli D., Fourestey G., Schmidt F. High-Throughput Lossy-to-Lossless 3D Image Compression. IEEE Transactions on Medical Imaging. 2021; 40(2): 607-620. DOI: 10.1109/TMI.2020.3033456.

9. Alcaín E., Fernández P. R., Nieto R. Hardware Architectures for Real-Time Medical Imaging. Electronics. 2021; 10(24): 3118-3126. DOI: 10.3390/electronics10243118.

10. Escande P., Weiss P. Fast wavelet decomposition of linear operators through product-convolution expansions. IMA Journal of Numerical Analysis. 2022; 42(1): 569-596. DOI: 10.1093/imanum/draa072.

11. Semenov V. I., Chumarov S. G. From wavelet construction based on Gaussian derivatives to finite impulse response filter synthesis. Nauchno-tekhnicheskii vestnik informatsionnykh tekhnologii, mekhaniki i optiki = Scientific and Technical Bulletin of Information Technologies, Mechanics and Optics. 2024; 24(2): 306-313. (In Russ.). DOI: 10.17586/2226-1494-2024-24-2-306-313.

12. Lyakhov P. A., Nagornov N. N., Semenova N. F., Abdulsalyamova A. Sh. Development of algorithms for digital image processing based on the Vinograd method in general and analysis of their computational complexity. Komp'yuternaya optika = Computer Optics. 2023; 47(1): 68-78. (In Russ.). DOI: 10.18287/2412-6179-CO-1146.

13. Ivanov V. E., Chye E. U. Modular discrete-analog wavelet filters. Moscow: KnoRus; 2021. 168 p. (In Russ.).

14. Bergerman M. V. Using a residual class system with modules of the form {2n - 1, 2n, 2n + 1} to reduce the hardware costs of a digital filter. Izvestiya vysshikh uchebnykh zavedenii. Povolzhskii region. Tekhnicheskie nauki = News of higher educational institutions. Volga region. Technical sciences. 2023; (1): 32-43. (In Russ.). DOI: 10.21685/20723059-2023-1-3.

15. Lavin A., Gray S. Fast algorithms for convolutional neural networks. Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 4013-4021.

16. Mehrabian A., Miscuglio M., Alkabani Y., Sorger V. J., El-Ghazawi T. A winogradbased integrated photonics accelerator for convolutional neural networks. IEEE Journal of Selected Topics in Quantum Electronics. 2019; 26(1): 1-12. DOI: 10.1109/JSTQE.2019.2957443.

17. Shen J., Huang Y., Wen M., Zhang C. Toward an efficient deep pipelined templatebased architecture for accelerating the entire 2-D and 3-D CNNs on FPGA. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems. 2019; 39(7): 14421455. DOI: 10.1109/TCAD.2019.2912894.

18. Nagornov N. N. Determination of the minimum bit depth of wavelet filter coefficients for three-dimensional medical imaging. Informatsionnye tekhnologii = Information Technologies. 2021; 27(8): 425-434. (In Russ.). DOI 10.17587/it.27.425-434.

19. Winograd S. Arithmetic complexity of computations. Siam, 1980; 33.

20. Chervyakov N., Lyakhov P., Kaplun D., Butusov D., Nagornov N. Analysis of the quantization noise in discrete wavelet transform filters for image processing. Electronics. 2018; 7(8): 135. DOI: 10.3390/electronics7080135.

21. Parhami B. Computer arithmetic. New York, NY: Oxford university press; 2010; 20.