Parameter identification of point spread function in noisy and blur images

In consideration of the motion blur and defocus blur of images, a robust bispectrum-based method is proposed to estimate the blurred parameters in noisy images. Firstly, the bispectrum of a blurred standard test image is calculated. Then, the curve fitting is used to obtain the functional relationship between the statistical characteristics in bispectrum and the degraded parameters. Finally, the BP neural network trained by the above mentioned functional relations can identify the parameters in point spread function in other noisy images. The experimental results show that the proposed method is effective for defocus and motion images in certain ranges. When SNR is 25 dB, the obtained tolerance of blurred parameters is less than 0.5 pixel.