Parallel implementation of local thresholding in Mitrion-C

Image processing algorithms are often implemented in reconfigurable devices (FPGA) (Wiatr, 2003; Lai et al., 2007), which facilitate higher implementation performance and enable real-time data processing. Currently, several main development directions of image processing, analysis and recognition systems based on reconfigurable hardware can be indicated. Some designs focus on implementing image processing operations which have not been accomplished on the FPGA platform yet (Kokufuta and Maruyama, 2009). New design methodologies of image processing algorithms are proposed (Plavec et al., 2009), while existing algorithms are accelerated by implementing computing intensive routines in FPGA resources (Sotiropoulos and Papaefstathiou, 2009). Comparison of speed-up factor for various implementation platforms, i.e., GPU, CPU (GPP) and the FPGA, is also considered (Asano et al., 2009; Claus et al., 2009). Stereovision (Ibarra-Manzano et al., 2009), self-organizing (Wildermann et al., 2009) and other complex algorithms are also implemented on the FPGA. Furthermore, power reduction and power saving techniques in image processing systems are discussed (Kalaycioglu et al., 2009). An interesting image processing and analysis system which takes advantage of partial run-time reconfiguration is presented by Canto et al. (2009). Most of the algorithms are designed in hardware description languages: the VHDL and Verilog (Cho et al., 2007). However, in recent years several tools have been developed which enable creation of higher abstraction level designs and their subsequent transformation into hardware circuitry descriptions. Examples include Handle-C (Vitabile et al., 2004), Impulse-C (ImpulseC, 2009) and graphical tools like System Generator (Murthy et al., 2008), PICO Extreme (Denolf et al., 2009) and PixelStreams (Jabłoński et al., 2006). A broad overview of high-level tools for circuit design can be found in the works of Araby et al. (2007) and Edwards (2006). This work’s primarily focus is to verify the usefulness of the Mitrion-C language in image processing operations. The only former example of implementing image processing algorithms in Mitrion-C is the Sobel edge detection, as presented in MitrionUserGuide (2008). Originally, only simple video channel and LUT (look-up table) operations were implemented (Sections 3 and 4), followed by implementation of two variants of the local thresholding algorithm—simple and Sauvola’s (Section 5), the latter requiring real number computations. Fixed-point and floating-point implementations were compared (Section 5.4). Additionally, a modification to Sauvola’s algorithm was proposed, enabling a reduction of FPGA resource usage.