Lattice architectures for multiple-scale gaussian convolution, image processing, sinusoid-based transforms and gabor filtering
Tóm tắt
This article describes a novel lattice architecture which performs multiple-scale Gaussian convolution of signals of any desired dimension. The principle of operation is based on the central limit theorem and involves repetitive convolution with small kernels to generate Gaussian smoothing. A pyramidal implementation of this architecture is also discussed which retains the fine scale (standard deviation) resolution, while reducing the complexity of the architecture. The lattice generates the scale-space description of the input signal which is very useful in image processing and vision applications. Applications of the lattice for Canny's and Laplacian of Gaussian edge detection are outlined. Using a novel Gaussian approximation to sinusoid kernels, the lattice is also capable of generating discrete Fourier, cosine, and sine transforms. Since the lattice can perform Gaussian smoothing and generate sinusoid kernels, it easily performs Gabor filtering and also reconstructs signals from their Gabor (or Gaussian) representations. An adaptive configuration of the lattice has been used for a hardware real-time implementation for our recently developed generalized nonorthogonal signal representations by Gaussians and Gaussian set wavelets.
Từ khóa
Tài liệu tham khảo
J. Ben-Arie, “Multidimensional linear lattice for Fourier and Gabor transforms, multiple-scale Gaussian filtering and edge detection,” inNeural Networks for Human and Machine Perception (H. Wechsler, ed.), Academic Press: New York, pp. 231–252, 1992.
A.P. Witkin, “Scale-space filtering: a new approach to multiscale description,” inImage Understanding 1984 (S. Ullman and W. Richards, eds.), Ablex: New York, 1984.
A. Papoulis,Probability, Random Variables, and Stochastic Processes, McGraw-Hill: New York, 1989.
W.M. Wells, “Efficient synthesis of Gaussian filters by cascaded uniform filters,”IEEE Trans. Patt. Anal. Machine Intell., vol. PAMI-8, No. 2, pp. 234–239, 1986.
J. Ben-Arie and K.R. Rao, “Signal representation by generalized non-orthogonal Gaussian wavelet groups using lattice networks,” inProc. IEEE Int. Joint Conf. Neural Networks, Singapore, pp. 968–973, Nov., 1991.
J. Ben-Arie and K.R. Rao, “Novel transform for image description and compression with implementation by neural architectures,” inProc. SPIE Conf. Stochastic and Neural Methods in Signal Processing, Image Processing and Computer Vision, San Diego, pp. 356–366, 1991.
J. Ben-Arie, “Linear lattice architectures that utilize the central limit for image analysis, Gaussian operators, sine, cosine, Fourier and Gabor transforms,” inProc. IEEE and SPIE Conf. Visual Communications and Image Processing '91, Boston, MA., pp. 823–838, Nov., 1991.
J. Ben-Arie, “Parallel generation of Fourier and Gabor transforms and other shape descriptors by Gaussian wavelet sets, using a set of multidimensional lattices,” inProc. IEEE 7th Workshop Multidimensional Signal Processing, Lake Placid, NY, Sec. 10.3, Sep., 1991.
J. Ben-Arie and K.R. Rao, “Lattice architectures for non-orthogonal representation of signals and generation of transforms using Gaussian sets,” Technical Report IIT ECE-TR-005-92, 1992.
J. Ben-Arie, “Image processing operators and transforms generated by a set of neural lattices that use the central limit,” inProc. IEEE Int. Joint Conf. Neural Networks, Singapore, pp. 987–993, Nov., 1991.
J. Ben-Arie and K.R. Rao, “A lattice network for signal representation using Gaussian basis functions and max-energy paradigm,” inProc. IEEE 34th Midwest Symp. Circuits and Systems, Monterey, CA, pp. 76–79, May, 1991.
P.J. Burt and E.H. Adelson, “The Laplacian pyramid as a compact image code,”IEEE Trans. Comm., vol. COM-31, pp. 532–540, 1983
P.J. Burt, “Fast filter transforms for image processing,”Computer Graphics Image Process., vol. 16, 20–51, 1981.
J. Canny, “A computational approach to edge detection,”IEEE Trans. Patt. Anal. Mach. Intell., vol. PAMI-8, no. 6, 1986.
C. Koch, “Seeing chips: Analog VLSI circuits for computer vision,”Neural Comput. vol. 1, no. 2, pp. 184–200, 1989.
J.P. Sage,” Gaussian convolution of images stored in a chargecoupled device,” Quarterly Technical Report, Aug. 1–Oct. 31, 1983, pp. 53–59, MIT Lincoln Laboratory, Lexington, MA, 1983.
T.F. Knight, Jr., “Design of an integrated optical sensor with on-chip preprocessing,” Ph.D dissertation, M.I.T., Department of Electrical Engineering and Comp. Science, 1983.
H. Kobayashi, J.L. White, and A.A. Abidi, “An active resistor network for Gaussian filtering of images,”IEEE Journal of Solid-State Circuits, vol. 26, no. 5, pp. 738–738, 1991.
A. Lumsdaine, J.L. Wyatt, and I.M. Elfadel, “Non-linear analog networks for image smoothing and segmentation,”J. VLSI Signal Process., no. 3, pp. 56–68, 1991.
J. Oyster, F. Vicuna, and W. Broadwell, “Associative network applications to low-level machine vision,”Appl. Opt., vol. 26, no. 10, pp. 1919–1926, 1987.
A.D. Culhane, M.C. Peckerar, and C.R.K. Marian, “A neural net approach to discrete Hartley and Fourier transforms,”IEEE Trans. Circuits Systems, vol. 36, no. 5, pp. 695–702, 1989.
J.G. Daugman, “Complete discrete 2D Gabor transform by neural networks for image analysis and compression,”IEEE Trans. ASSP, vol. 36, no. 7, pp. 1169–1179, 1988.
W. Feller,An Introduction to Probability Theory and Its Applications, Wiley: New York, 1957.
D. Marr,Vision, Freeman: New York, 1982.
M. Porat and Y.Y. Zeevi, “The generalized Gabor scheme of image representation in biological and machine vision,”IEEE Trans. PAMI, vol. 10, no. 4, pp. 452–468, 1988.
T. Ebrahimi and M. Kung, “Image compression by Gabor expansion,”Opt. Eng., vol. 30, no. 7, pp. 873–880, 1991.