350 rub
Journal Science Intensive Technologies №7 for 2013 г.
Article in number:
Obtaining of light spatial noise portrait of camera-s photosensor for digital cameras identification and increasing of signal-to-noise ratio at image registration
Keywords: digital camera photosensor signal-to-noise ratio spatial noise light spatial noise portrait camera identification
Authors:
N.N. Evtikhiev, S.N. Starikov, P.A. Cheryomkhin, O.A. Yatsushko
Abstract:
In the article method of obtaining of light spatial noise portrait of digital camera photosensor is presented and experimentally tested. This method yields greater accuracy for account of iterative process of creating luminance with high spatial homogeneity and improved technique of temporal and spatial dark noises suppression. By using the proposed method light spatial noise portrait of camera Canon EOS 400D was obtained. The relative error of the portrait elements was 15-20%. Using the obtained portrait, an experiment for source camera identification was conducted. Among set of snapshots taken by various chambers, all snapshots taken by predetermined digital camera were correctly identified. Using the only blue color channel, a minimum ratio of the correlation coefficients own / another is at least 60.0. Analytical estimates of increasing of signal/noise ratio (SNR) at image registration with spatial noise compensation were derived. Estimates of SNR were obtained by single and multiple exposures. To test the consistency of the obtained estimates, the numerical simulation of image registration with camera Canon EOS 400D was conducted. The results of the numerical simulation are in good agreement with the calculated estimates. The results of numerical experiments are in good agreement with the analytical estimates. Analytical estimates and results of numerical experiments show that the use of spatial noise compensation with frames averaging can provide a significant increase in SNR. With compensation of accurate spatial noise portrait, increasing of SNR can be expected to 50 times compared to a single image. In comparison with the case of only temporal noise suppression, increasing of SNR can be expected up to 50 times. With compensation of approximate spatial noise portrait (20% relative error), increasing of SNR can be expected to 5 times compared to a single image. In comparison with the case of only temporal noise suppression, increasing of SNR can be expected up to 10 times. In order to achieve SNR value, which is close to the maximum, averaging of 300-500 shots is enough.
Pages: 55-64
References

  1. Janesick J. Scientific Charge-Coupled Devices. Bellingham-Washington: SPIE Press. 2001.
  2. Lukáš J., Fridrich J., Goljan M. Digital Camera Identification from Sensor Pattern Noise // IEEE Transactions on Information Security and Forensics. 2006. V. 1. № 2. P. 205-214.
  3. Lukáš J., Fridrich J., Goljan M. Detecting Digital Image Forgeries Using Sensor Pattern Noise // Proc. SPIE. 2006. V. 6072. P. 0Y1-0Y11.
  4. Fienup J.R. Invariant error metrics for image reconstruction // Appl. Opt. 1997. V. 36. №32. P. 8352-8357.
  5. Nakamura J. Image sensors and signal processing for digital still cameras. Boca Raton, FL: CRC Press. 2006.
  6. The European Machine Vision Association. EMVA Standard 1288, Standard for Characterization of Image Sensors and Cameras, 2010. URL: http://www.emva.org/cms/upload/Standards/Stadard_1288/EMVA1288-3.0.pdf (Data obrashheniya: 8.02.2013).
  7. Konnik M.V., Many'kin E'.A., Starikov S.N. Rasshirenie vozmozhnostej kommercheskoj czifrovoj fotokamery' dlya registraczii prostranstvenny'x raspredelenij intensivnosti lazernogo izlucheniya // Kvantovaya e'lektronika. 2010. T. 40. №4. S. 314-320.
  8. Evtixiev N.N., Starikov S.N., Cheryomxin P.A. Oczenka vliyaniya dinamicheskogo diapazona i shumov registriruyushhix kamer na kachestvo czifrovy'x gologramm // Opticheskij zhurnal. 2013. T. 80. №7.