An identification method of blur parameter in noisy and mixed blur images

There are motion blur and defocus blur in the images captured by optics cameras in some areas, for example, in the areas of aeroplane photography and navigation using scene matching. Noise makes parameter identification of point spread function more difficult in mixed blur images. A robust bispectrum-based method to estimate the blur parameter in noisy and mixed blur images is proposed. At first, two bispectrums are calculated according to motion blur and defocus blur of a standard test image. Then the curve fitting is used to get the function relations between statistics characteristic in two bispectrums and degraded parameters. Finally, the BP neural network, which is trained by the aforementioned function relations, can accomplish the identification of blur parameter in other noisy and mixed blur images. The experimental results show that the method is effective in certain ranges. When SNR is 25dB, the tolerance of blur parameters recognized is less than 0.5 pixel.