Skip to main content

Table 3 Most relevant results of the retinal vessel tree extraction algorithm implementation on the different devices

From: Performance analysis of massively parallel embedded hardware architectures for retinal image processing

 

Intel Core i7 940

SCAMP-3

Spartan-3

Am2045

Window size (px)

-

128 × 128

144 × 160

-

Processors (used/available)

4/4

16,384/16,384

90/90

125/360

Working frequency (MHz)

2930

1.25

53

333

Window execution time (ms)

-

6.55

66.1

-

Required windows

1

30

20

1

Computation time (s)

13.6

0.193

1.323

0.008

Total execution time with I/O (s)

13.7

0.230

1.349

0.0087

Speed-up

59.6×

10.2×

1574.7×*

Cycles-per-pixel (without I/O)

357,993

8,950

14,070

742*

  1. * It should be noted that the Ambric Am2045 runs a simplified version of the algorithm. See Section 6 for details