In this work, we present a FPGA-based Generalized Hough Transform custom processor to calculate similarities between arbitrary shapes. Raw data are 44x36 DC images extracted directly from low-resolution compressed video (352x288). The outputs are two numbers per frame that quantify the image similitude in terms of scale and rotation. The proposed architecture efficiently resolves the detection of pixel pairs, and the voting of distances and rotations, without memory access conflicts. These operations are inherent to Hough transformation. The paper condenses some circuit solutions suitable to hardwiring video processing. They take full advantage of using small embedded memories as look-up tables. The complete processor is validated with benchmark video samples that cover different scenarios and problems: sport, drama, and news. The final version internally operates at 100 MHz and fits inside a small FPGA chip. The highly concurrent architecture employs both pipelining and parallelism using hardware replication. The final performance is over 40 Giga fixed-point operations per second.
This paper is a study of the Nios II power characterization. The relationship between instruction and data cache sizes and the corresponding energy consumption is analyzed. The study is based on more than one thousand current measurements for different benchmark programs and cache configurations. From the results it is clear that using the optimal cache sizes leads to the lowest energy consumption even when considering execution time, power, and FPGA resources utilization. As an additional result the paper shows an example where the use of integer instead of floating point operations can save a significant amount of energy. Lastly, it is shown that the energy consumption as a function of the input data size follows the same function as the computational complexity for the studied examples.
Compton cameras acting as electronic collimators, improve the characteristics of nuclear medicine imaging, with an additional computational cost for the readout circuitry and digital control system. This paper presents the use of reconfigurable logic based hardware-software co-design, to deal with such tasks as the ones related to the electronic collimation architecture used on gamma cameras. The readout circuitry has been implemented as dedicated logic and as an embedded system that contains it. Both are studied in terms of scalability in the number of channels and response time. This number of channels and the speed of acquisition of such channels define the overall performance achieved by the technique. A typical scenario of electronic collimation application, using a high purity pixelated CdZnTe crystal as the sensitive material with a pixellated anode design, is the base of this performance study. The results of this study are the exact design parameters of an FPGA based embedded system that performs all tasks associated with the control of an electronic collimator prototype.
This paper shows that, under certain conditions, digital arithmetical circuits do not meet the addition commutation property in terms of power consumption. That is, the power consumed by the operation AtimesB is different from BtimesA. As a consequence, it is possible to get a power saving simply permuting the circuit inputs, wherever any of the following three conditions are present: a) the data to be processed has a strong temporal correlation; b) the delays between the circuit paths are highly unequalized; c) one of the input data communication is broadcast type, meanwhile the other is local. In order to verify these hypotheses, several binary multipliers were constructed and measured. The power consumption reduction resulted between 12% and 28% in Virtex FPGAs.