Single image motion deblurred based on fractional differential operator

摘要

This paper proposes a motion deblurred method of single image based on fractional differential operator. According to the periodicity of motion blurry image in frequency domain, the proposed method estimates the Point Spread Function of image based on self-correlation in blurry image and twice Fourier transform in gradient image. After the Point Spread Function is estimated, we estimate the latent image by adding the fractional differential operator as the regularization term. Compared with traditional image deblurred methods, the proposed method makes a trade-off between over smooth and sharpen since fractional differential operator can preserve textures and restrain noise. Experimental results show that the restored image has better Peak Signal to Noise Ratio and Structural Similarity Index than other image deblurred methods.