B.J. Falkowski

Nanyang Technological University, Tumasik, Singapore

Are you B.J. Falkowski?

Claim your profile

Publications (255)41.76 Total impact

  • Electrical Review. 01/2010; 11a:95-101.
  • Source
    [show abstract] [hide abstract]
    ABSTRACT: With the evolution of programmable structures, that become more heterogeneous, the process of mapping a design into these structures becomes more and more complex. Modern FPGA chips are equipped with embedded memory blocks that can be used to increase the implementation quality of the design. The paper presents a logic synthesis method based on balanced decomposition that uses the concept of r-admissibility to efficiently utilize possibilities provided by memory blocks embedded in modern FPGA architectures. Results presented in this paper prove the effectiveness of proposed approach.
    Mixed Design of Integrated Circuits & Systems, MIXDES '09. MIXDES-16th International Conference, Lodz, Poland; 06/2009
  • Source
    Grzegorz Borowik, Tadeusz Luba, Bogdan J. Falkowski
    [show abstract] [hide abstract]
    ABSTRACT: This paper presents a new cost-efficient realization scheme of pattern matching circuits in FPGA structures with embedded memory blocks (EMB). The general idea behind the proposed method is to implement combinational circuits using a net of finite state machines (FSM) instead. The application of functional decomposition method reduces the utilization of resources by implementing FSMs using both EMBs and LUT-based programmable logic blocks available in contemporary FPGAs. Experimental results for the proposed method are also shown. A comparison with another dedicated method yields extremely encouraging results: with a comparable number of EMBs, the number of logic cells has been reduced by 95%.
    12th IEEE Symposium on Design and Diagnostics of Electronic Circuits and Systems, DDECS 2009, Liberec, Czech Republic; 04/2009
  • Source
    Bogdan J. Falkowski, Sujit Kumar Sahoo, T. Luba
    [show abstract] [hide abstract]
    ABSTRACT: This paper presents two novel methods for compression of fluorescent dye cell images. The compression methods achieve high compression ratio by pseudo-quantizing the pixel gray levels according to image statistics and extracting and saving the useful regions of the image. The proposed algorithms produce a better compression over the existing lossless image formats in the case of fluorescent dye cell images and this idea of compression can be further extended to other images.
    01/2009;
  • Cicilia C. Lozano, Bogdan J. Falkowski, T. Luba
    [show abstract] [hide abstract]
    ABSTRACT: New quaternary linearly independent arithmetic (QLIA) transforms are presented in this paper. The new transforms are recursive and they are classified into two classes according to their recursive equations. Fast flow graphs and relations for the transforms are presented. Computational cost for the calculation of the QLIA spectral coefficient vector is also derived. Experimental results in terms of the minimum number of nonzero spectral coefficients based on the QLIA transforms for a set of quaternary test functions are also given and compared with the corresponding numbers for quaternary fixed polarity arithmetic (QFPA) transforms. The comparison shows that for the set of quaternary test functions the average number of minimum number of nonzero spectral coefficients based on the introduced QLIA transforms is 39.28% and 16.29% smaller than that for the polarity zero and optimal polarity QFPA transforms, respectively.
    01/2009;
  • Cicilia C. Lozano, Bogdan J. Falkowski, Tadeusz Luba
    [show abstract] [hide abstract]
    ABSTRACT: A fixed polarity quaternary linearly independent (FPQLI) transform is introduced in this paper where the basic transforms are derived from the recursive structure of some linearly independent transforms over Galois Field (2) (GF(2)). For some polarities the FPQLI transform for n-variable quaternary functions directly corresponds to the binary fixed polarity Reed-Muller (FPRM) transforms for 2n-variable binary functions. In this paper, the fast flow graph and recursive equations for the FPQLI transform are given together with the underlying basis functions. Formulae for converting the FPQLI spectral coefficient vector from one polarity to another are also given and used to generate a recursive algorithm to obtain the optimal FPQLI expansion with reduced computational cost. Experimental results of the FPQLI transform have been obtained for a set of quaternary test files. Comparison of the obtained results with FPRM over GF(4) as well as the existing recursive quaternary linearly independent transforms show the advantage of applying the concept of fixed polarity to the basic transforms for function minimization in terms of smaller number of nonzero spectral coefficients.
    ISMVL 2009, 39th International Symposium on Multiple-Valued Logic, 21-23 May 2009, Naha, Okinawaw, Japan; 01/2009
  • C. C. Lozano, B.J. Falkowski, T. Luba
    [show abstract] [hide abstract]
    ABSTRACT: Spectral expansions are alternative representations of logic functions/signals in which the information are redistributed and presented differently in terms of spectral coefficients. The use of spectral representations often allows certain operations or analysis to be performed more efficiently on the data. In this paper, spectral expansions for ternary functions based on new fastest linearly independent arithmetic transforms are presented and discussed. The new transforms are generalizations of some existing ternary transforms through permutation and reordering operations. They have regular structures and can be computed using fast transforms. Formulae for their fast forward and inverse transformations as well as their corresponding fast flow graphs are shown here. Computational costs and some properties of the transforms and their spectra are also given. Finally, experimental results of the transforms are presented which show that the new transforms can represent some functions more compactly than the existing transforms.
    Mixed Design of Integrated Circuits and Systems, 2008. MIXDES 2008. 15th International Conference on; 07/2008
  • B.J. Falkowski, Cheng Fu
    [show abstract] [hide abstract]
    ABSTRACT: Classification of novel fastest quaternary linearly independent transforms has been presented. They are recursively defined and have consistent formulas relating their forward and inverse transform matrices. Their transform matrices' properties and calculation example have been shown. The computational costs of the calculation for presented transforms are also discussed. The experimental results are shown and compared with the well known quaternary arithmetic transform.
    Multiple Valued Logic, 2008. ISMVL 2008. 38th International Symposium on; 06/2008
  • C.C. Lozano, B.J. Falkowski, T. Luba
    [show abstract] [hide abstract]
    ABSTRACT: Sixteen fastest quaternary linearly independent (FQLI) transforms are discussed in this paper. All the presented transforms can be derived using recursive equations and their inverse recursive definitions share common structure. In this paper, the fast flow graph equations and properties for the FQLI transforms are given. Based on the relationships between the spectra of the transforms, a recursive algorithm for the computation of their spectral coefficients is also proposed which reduces the total computational cost of generating their complete spectra. Experimental results for all the FQLI transforms as well as fixed polarity Reed-Muller transform over GF(4) are also given and compared in terms of the number of nonzero spectral coefficients. The comparison shows that for some quaternary functions the new FQLI transforms can give more compact representations.
    Multiple Valued Logic, 2008. ISMVL 2008. 38th International Symposium on; 06/2008
  • M. Rawski, T. Luba, B.J. Falkowski
    [show abstract] [hide abstract]
    ABSTRACT: The paper presents a logic synthesis method oriented towards FPGA architectures with specialized embedded memory blocks. Existing methods do not ensure effective utilization of possibilities provided by these specialized embedded modules. The presented method, based on balanced decomposition, leads to much more effective implementations of digital systems in modern FPGA structures.
    Circuits and Systems, 2008. ISCAS 2008. IEEE International Symposium on; 06/2008
  • B.J. Falkowski
    [show abstract] [hide abstract]
    ABSTRACT: In this paper, a robust phase watermarking algorithm for still images is presented where the watermark information is conveyed in the phase spectrum in the transform domain. Phase watermarking algorithm that uses multi-polarity Complex Hadamard transform is developed. The robustness of presented algorithm is investigated by its uniqueness, JPEG encoding, image resizing, dithering noise distortions, sharpening, cropping and successive watermarking.
    Information, Communications & Signal Processing, 2007 6th International Conference on; 01/2008
  • C.C. Lozano, B.J. Falkowski
    [show abstract] [hide abstract]
    ABSTRACT: New fastest linearly independent ternary arithmetic (LITA) transforms are presented in this paper. Fastest LITA transforms can be calculated efficiently by fast transforms and the calculation incurs low computational complexity. Formulas for the fast forward and reverse transformation of the new fastest LITA transforms are given and several of their properties are listed. A modular tree structure hardware implementation of the resulting fastest LITA polynomial expansion is also described. At the end of the paper, experimental results for the new transforms are shown.
    Information, Communications & Signal Processing, 2007 6th International Conference on; 01/2008
  • Bogdan J. Falkowski
    [show abstract] [hide abstract]
    ABSTRACT: In this paper, a robust phase watermarking algorithm for still images is presented where the watermark information is conveyed in the phase spectrum in the transform domain. Phase watermarking algorithm that uses multi-polarity Walsh-Hadamard and Complex Hadamard transform is developed. The robustness of presented algorithm is investigated by its uniqueness, JPEG encoding, image resizing, dithering noise distortions, sharpening, cropping and successive watermarking.
    Journal of Mathematical Imaging and Vision 01/2008; 30:13-21. · 1.77 Impact Factor
  • Source
    Mariusz Rawski, Bogdan J Falkowski, Tadeusz Łuba
    [show abstract] [hide abstract]
    ABSTRACT: This paper presents the discussion on efficiency of different implementation methodologies of DSP algorithms targeted for modern FPGA architectures. Modern programmable structures are equipped with specialized DSP embedded blocks that allow implementing digital signal processing algorithms with use of the methodology known from digital signal processors. On the first place, however, programmable ar-chitectures give the designer the possibility to increase efficiency of designed system by exploitation of parallelism of implemented algorithms. Moreover, it is possible to apply special techniques, such as distributed arithmetic (DA) that will boost the per-formance of designed processing systems. Additionally, application of the functional decomposition based methods, known to be best suited for FPGA structures, allows utilizing possibilities of programmable technology in very high degree. The paper presents results of comparison of different design approaches in this area.
    01/2008; 20:437-459.
  • B. J. Falkowski, C. C. Lozano, T. Luba
    [show abstract] [hide abstract]
    ABSTRACT: Reliability analysis of a system conducted in the Probabilistic (Arithmetic) domain is lengthy and errorprone. An alternative method is to consider the system reliability components to be Boolean variables rather than Probabilistic variables and to treat the whole problem as if it were Boolean. This method allows the use of powerful Boolean reduction theorems to contain the size of the problem. The Boolean expression should be fully disjoint so that the conversion back to the Probabilistic domain would be on a one-to-one basis. Besides presenting basic properties and definitions of Probabilistic and Boolean domains, evaluation of reliability of a system directly from Arithmetic expansion is also discussed. Since the reliability expression may be directly obtained from disjoint cube representation, two methods are shown.
    01/2008;
  • Tao Fang, Cheng Fu, Bogdan J. Falkowski, Bang Wang
    [show abstract] [hide abstract]
    ABSTRACT: High dynamic range (HDR) images require a higher number of bits per color channel than traditional images. This brings about problems to storage and transmission, especially in distributed wireless sensor networks in which transmission capability and energy consumption are critical issues. The paper proposes a structure of compressing multiple dynamic range images, which is called multiple dynamic range (MDR) coding. MDR imaging system is attractive due to the capability of reconstructing lower dynamic range images from partial bitstream. MDR coding shows its advantages in terms of energy efficiency in distributed wireless sensor networks. The efficacy of MDR coding is also illustrated by presenting the results of encoding a series of synthetic and natural MDR images.
    01/2008;
  • B.J. Falkowski, Shixing Yan
    [show abstract] [hide abstract]
    ABSTRACT: The new ternary Walsh transform is considered in this paper. Such a ternary Walsh transform can be used in a similar manner as the standard Walsh transform for binary logic functions as shown here. It is based on the Kronecker product as well as the Galois field and new ternary operations. The same hardware implementation can be used for both forward and inverse ternary Walsh transforms based on its fast algorithms and properties. The ternary Walsh transform is suitable for processing both completely and incompletely specified Boolean functions. Its properties for the decomposition and symmetry detections of the Boolean functions are shown.
    Circuits and Systems I: Regular Papers, IEEE Transactions on 09/2007; · 2.24 Impact Factor
  • Source
    [show abstract] [hide abstract]
    ABSTRACT: General functional decomposition has been gaining more and more importance in recent years. Though it is mainly perceived as a method of logic synthesis for implementation of Boolean functions into FPGA-based architectures, it has found application in many other fields of modern engineering and science. In this paper, application of logic decomposition to efficient implementation of image and signal processing system in modern FPGA architectures is described. We have adapted an algorithm called Demain that can be used as a preprocessing tool in the FPGA synthesis flow. Wavelet filters are used to test Demain with Altera's Stratix device family and the experimental results are compared with those by Altera's Quartus tool. When Demain is used as a premapping processing, it shows 50% improvement in terms of the covered area by embedded memory blocks.
    Mixed Design of Integrated Circuits and Systems, 2007. MIXDES '07. 14th International Conference on; 07/2007
  • B.J. Falkowski, T. Sasao, T. Luba
    [show abstract] [hide abstract]
    ABSTRACT: This article shows an algorithm to generate Walsh functions in four different orderings: Hadamard, Harmuth, Paley and strict sequency. The properties and mutual relations among these four orderings are analyzed and a method is developed that allows any of the orderings to be generated from the primary set of Rademacher functions. Finally, a programmable Walsh function generator for 64 outputs is implemented using both field programmable gate arrays (FPGAs) and lookup table (LUT) cascades to estimate the amount of hardware and performance. Such a programmable Walsh function generator can be used in VLSI testing, CDMA, pattern recognition as well as image and signal processing.
    Mixed Design of Integrated Circuits and Systems, 2007. MIXDES '07. 14th International Conference on; 07/2007
  • C.C. Lozano, B.J. Falkowski, S. Rahardja
    [show abstract] [hide abstract]
    ABSTRACT: New fast linearly independent arithmetic (LIA) transforms are introduced here which can be used to represent any functions of binary variables. The transforms are grouped into classes where consistent formulas relating forward and inverse transform matrices are obtained. All the presented transforms have the same computational cost, which is lower than the computational cost of the well-known fixed polarity arithmetic transforms. General classifications and fast forward and inverse transform definitions for all the fast LIA transforms are given. Various properties and mutual relations that exist for the different transforms and their corresponding spectra are also shown. The presented relations and properties reduce the computational cost of finding the best LIA polynomial expansion based on the new transforms.
    IET Signal Processing 07/2007; · 0.71 Impact Factor

Publication Stats

941 Citations
41.76 Total Impact Points

Institutions

  • 1992–2009
    • Nanyang Technological University
      • School of Electrical and Electronic Engineering
      Tumasik, Singapore
  • 2007–2008
    • Warsaw University of Technology
      • Institute of Telecommunications
      Warsaw, Masovian Voivodeship, Poland
  • 1992–2007
    • Nanyang Normal University
      Nan-yang-shih, Henan Sheng, China
  • 2005
    • Institute for Infocomm Research
      Tumasik, Singapore
  • 1997
    • University of Niš
      Nisch, Central Serbia, Serbia
  • 1990–1997
    • Portland State University
      • Department of Electrical and Computer Engineering
      Portland, OR, United States