Error Detection Technique For A Median Filter: Luis Alberto Aranda, Pedro Reviriego, and Juan Antonio Maestro
Error Detection Technique For A Median Filter: Luis Alberto Aranda, Pedro Reviriego, and Juan Antonio Maestro
Error Detection Technique For A Median Filter: Luis Alberto Aranda, Pedro Reviriego, and Juan Antonio Maestro
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY CALICUT. Downloaded on March 09,2020 at 04:52:33 UTC from IEEE Xplore. Restrictions apply.
2220 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 64, NO. 8, AUGUST 2017
Fig. 2. (a) Original, (b) median filtered, and (c) bad filtered Lena image [10].
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY CALICUT. Downloaded on March 09,2020 at 04:52:33 UTC from IEEE Xplore. Restrictions apply.
ARANDA et al.: ERROR DETECTION TECHNIQUE FOR A MEDIAN FILTER 2221
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY CALICUT. Downloaded on March 09,2020 at 04:52:33 UTC from IEEE Xplore. Restrictions apply.
2222 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 64, NO. 8, AUGUST 2017
TABLE I
U TILIZATION R EPORT
Fig. 9. Standard test images. (a) Baboon. (b) Boat. (c) Cameraman.
(d) House. (e) Lena. (f) Peppers [10].
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY CALICUT. Downloaded on March 09,2020 at 04:52:33 UTC from IEEE Xplore. Restrictions apply.
ARANDA et al.: ERROR DETECTION TECHNIQUE FOR A MEDIAN FILTER 2223
TABLE II TABLE IV
AVERAGED FAULT-I NJECTION R EPORT (I MAGE -L EVEL ) AVERAGED E RROR D ETECTION R EPORT (P IXEL -L EVEL )
TABLE III
AVERAGED E RROR D ETECTION R EPORT (I MAGE -L EVEL ) In SRAM-based FPGAs, these false detections mean that an
undesirable reconfiguration is performed, so reprogramming
time and power are wasted. In addition to the waste of time,
the filter is not operational during the reconfiguration time, so
it can be critical depending on the application.
Table III also presents the MSE value of the undetected
images. MSE is calculated using (1), where Pi represents the
original uncorrupted median pixel values, and P i the observed
pixel values
N−1
MSE = [Pi − P i ]2 . (1)
i=0
capabilities. In order to illustrate the differences between these Our proposed scheme presents the lowest averaged MSE
three methods, three tests have been performed. First, in value of the three studied techniques. This means that the
Section V-A, the fault injection results have been summarized corrupted pixel values of the undetected images are closer
in image-level reports. Meaning that the configuration memory to the uncorrupted golden image, so these corrupted pixels
bit flip is performed and then the entire image in pixel- can be used by the following image processing steps without
stream format is processed by the median filter. If at least significantly altering the final results.
one erroneous pixel is detected, then the image is classified
as corrupted but detected. In Section V-B, these image-level B. Pixel-Level Results
reports are subdivided collecting the individual erroneous pixel
A pixel-level report summarizes the total percentage of
values per image. This way, the total number of corrupted but
corrupted pixels detected. This additional information is useful
detected pixels can be determined. Finally, in Section V-C,
to understand how each technique works at low-level. Table IV
the input test images are analyzed to prove that they cover a
presents the average number of corrupted pixels detected
wide variety of input scenarios. The purpose of this analysis
by each technique. As can be observed in this table, the
is to test that the error detection capabilities of the studied
proposed implementation detects 59% of the total number
techniques are not dependent of the inputs.
of pixels in average, however, as was observed in Table III,
it also detects 91% of corrupted images. This is because one
A. Image-Level Results corrupted pixel detected is enough to classify an image as
An image-level report gives us information about the per- corrupted but detected. Here lies the difference with DMR.
centage of corrupted/uncorrupted images detected/undetected As mentioned before, DMR consists of two fully redundant
by the technique. The eighteen image-level error reports have copies of the original design for the purpose of detecting any
been summarized in Tables II and III. small change in the design. This implies a high detection
In Table II, the average number of corrupted and uncor- rate but also a high resource usage. Conversely, our detection
rupted images is presented for each technique. It should be technique is optimized in terms of resource usage to detect
noticed that the number of injections needed to cover the some particular cases when the median value is outside the
DMR scheme is higher due to its bigger resource usage. limits of a dynamic range.
Table III shows the error detection rate and the mean-square Our technique cannot detect those cases when the created
error (MSE). It can be observed that the proposed technique dynamic range is large or the changes in the median value are
detects 91% of the corrupted images, while the RPR scheme, small, because the median value still remains within the range.
due to its reduced precision, only achieves a detection rate In order to evaluate this behavior, the detected/undetected pixel
of 48%. This 91% of detection rate is still far from the values have been split into LSBs and MSBs changes in the
expected 99% of the DMR, however, our method presents median value. This is, the configuration memory error has
less number of false positives, which means that less detected modified the median value in less or equal than 15 (four LSBs)
errors occur when no design malfunction is observed. The or in more or equal than 16 (four MSBs changes). Table V
number of false positives is related with the resource overhead shows this pixel classification.
added to the original design. Therefore, the bigger the error As can be observed in this table, almost all the corrupted
detection branch, the higher the number of false positives. pixels detected by the RPR technique are those whose value
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY CALICUT. Downloaded on March 09,2020 at 04:52:33 UTC from IEEE Xplore. Restrictions apply.
2224 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 64, NO. 8, AUGUST 2017
TABLE V
AVERAGED E RROR D ETECTION C LASSIFICATION (P IXEL -L EVEL )
TABLE VI
AVERAGED N UMBER OF C ORRUPTED P IXELS PER I MAGE
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY CALICUT. Downloaded on March 09,2020 at 04:52:33 UTC from IEEE Xplore. Restrictions apply.
ARANDA et al.: ERROR DETECTION TECHNIQUE FOR A MEDIAN FILTER 2225
Fig. 16. Percentage of detected images for each technique and test image.
Fig. 14. Detection rates and false positives comparisons between Stars image
and the averaged case.
TABLE VIII
S TARS I MAGE E RROR D ETECTION C LASSIFICATION (P IXEL -L EVEL )
Fig. 17. Percentage of false positives detections for each technique and test
image.
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY CALICUT. Downloaded on March 09,2020 at 04:52:33 UTC from IEEE Xplore. Restrictions apply.
2226 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 64, NO. 8, AUGUST 2017
images analysis proves that the selected images cover a wide [2] B. Jahne, “Applications and tools,” in Digital Image Processing, 6th ed.
variety of scenarios and that the error detection rate of the Berlin, Germany: Springer, 1991, ch. 1, sec. 3, pp. 14–16.
[3] R. Szeliski, “Image processing,” in Computer Vision: Algorithms
proposed technique is not affected by the input images. and Applications. London, U.K.: Springer, 2010, ch. 3, secs. 2–3,
pp. 111–127.
VI. C ONCLUSION [4] L. Sterpone, M. S. Reorda, M. Violante, F. L. Kastensmidt, and
L. Carro, “Evaluating different solutions to design fault tolerant systems
This paper is an extension of our previous work presented in with SRAM-based FPGAs,” J. Electron. Test., vol. 23, no. 1, pp. 47–54,
[10] about a fault-tolerant implementation of the median filter. 2007.
[5] G. R. Hopkinson, “Radiation effects in a CMOS active pixel sensor,”
Our proposed technique checks if the median output value is IEEE Trans. Nucl. Sci., vol. 47, no. 6, pp. 2480–2484, Dec. 2000.
within a dynamic range created each time with the remaining [6] M. Bagatin, “Introduction to the effects of radiation on electronics
nonmedian outputs. devices,” in Ionizing Radiation Effects in Electronics, 1st ed. Boca Raton,
FL, USA: CRC Press, 2016, ch. 1, secs. 1–4, pp. 15–16.
The proposed technique has been compared in terms of [7] A. Bovik, Handbook of Image and Video Processing. New York, NY,
resource utilization and error detection rate with DMR and USA: Academic, 2000.
RPR schemes. The design has been implemented in a Xilinx [8] B. Shim and N. R. Shanbhag, “Reduced precision redundancy for
low-power digital filtering,” in Proc. Conf. Rec. 35th Asilomar Conf.
SRAM-based FPGA and several fault-injection campaigns Signals, Syst. Comput., vol. 1. Pacific Grove, CA, USA, Sep. 2001,
have been performed. Image-level experimental results show pp. 148–152.
that, adding 35% of FPGA resources to the original design, [9] P. Reviriego, J. A. Maestro, I. López, and J. A. de Agapito, “Soft error
tolerant infinite impulse response filters using reduced precision repli-
91% of the corrupted images can be detected. This lower over- cas,” in Proc. 12th Eur. Conf. Radiat. Effects Compon. Syst. (RADECS),
head means that a lower number of false positive detections Sevilla, Spain, Sep. 2011, pp. 493–496.
are performed, so less reprogramming time is wasted and, [10] L. A. Aranda, P. Reviriego, and J. A. Maestro, “A fault-tolerant imple-
consequently, the time that the median filter is not operational mentation of the median filter,” presented at the Conf. Radiat. Effects
Compon. Syst. (RADECS), Bremen, Germany, 2016.
due to the reconfiguration is minimized. [11] L. Yin, R. Yang, M. Gabbouj, and Y. Neuvo, “Weighted median filters:
Pixel-level results show that images with homogeneous A tutorial,” IEEE Trans. Circuits Syst. II, Analog Digit. Signal Process.,
pixel-value regions are less susceptible to be corrupted when vol. 43, no. 3, pp. 157–192, Mar. 1996.
[12] M. A. Vega-Rodríguez, J. M. Sánchez-Pérez, Juan A. Gómez-Pulido,
processed by a configuration memory damaged median filter. “An FPGA-based implementation for median filter meeting the
More homogeneous pixel-value regions imply more median real-time requirements of automated visual inspection systems,” in
filter inputs with equal values, so the probability that a median Proc. 10th Medit. Conf. Control Autom., Lisbon, Portugal, 2002,
pp. 1–7.
filter malfunction alters the median value is decreased. [13] K. Benkrid, D. Crookes, and A. Benkrid, “Design and implementation
The test images have also been analyzed to prove that of a novel algorithm for general purpose median filtering on FPGAs,”
they cover different input scenarios. These experiments have in Proc. IEEE Int. Symp. Circuits Syst. (ISCAS), vol. 4. May 2002,
pp. IV-425–IV-428.
revealed that the RPR method is dependent on the input [14] G. L. Bates and S. Nooshabadi, “FPGA implementation of a
images, while the detection rate of our proposed technique median filter,” in Proc. IEEE Region 10th Annu. Conf. Speech
is not affected by the input images. Image Technol. Comput. Telecommun. (TENCON), vol. 2. Dec. 1997,
pp. 437–440.
In conclusion, although our technique cannot detect as many [15] S. A. Fahmy, P. Y. K. Cheung, and W. Luk, “Novel FPGA-based imple-
corrupted pixels as the DMR scheme, it detects enough to mentation of median and weighted median filters for image processing,”
prevent 91% of the corrupted images from being erroneously in Proc. Int. Conf. Field Program. Logic Appl., Tampere, Finland,
Sep. 2005, pp. 142–147.
processed by the next image processing module. Moreover,
[16] K. E. Batcher, “Sorting networks and their applications,” in Proc.
the remaining 9% of undetected images presents a lower AFIPS Spring Joint Computing Conf., San Francisco, CA, USA, 1968,
MSE value than the DMR, which means that the image pp. 307–314.
damage propagation to successive image processing operations [17] Z. Vasicek and L. Sekanina, “Novel hardware implementation of adap-
tive median filters,” in Proc. 11th IEEE Workshop Design Diagnostics
is minimized. Electron. Circuits Syst. (DDECS), Bratislava, Slovakia, Apr. 2008,
pp. 1–6.
[18] J. L. Smith, “Implementing median filters in XC4000E FPGAs,” XCell,
vol. 23, p. 16, Jul. 1996.
R EFERENCES
[19] Zynq-7000 All Programmable SoC Overview DS190 V1.9, Xilinx,
[1] H.-L. Eng and K.-K. Ma, “Noise adaptive soft-switching median San Jose, CA, USA: Jan. 2016.
filter,” IEEE Trans. Image Process., vol. 10, no. 2, pp. 242–251, [20] Soft Error Mitigation Controller LogiCORE IP Product Guide (PG036),
Feb. 2001. Xilinx, San Jose, CA, USA: Sep. 2015.
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY CALICUT. Downloaded on March 09,2020 at 04:52:33 UTC from IEEE Xplore. Restrictions apply.