4.2 Image Processing in Multi-tag Movement
119
Fig. 4.10 Effect of PSF on images in the frequency domain
represents the distribution of the ideal picture in this coordinate, and the curve
log|HU | represents the distribution of the image after it interacts with the blur kernel
(PSF). This means that the effect of PSF on pictures has obvious characteristics in
the frequency domain.
Different estimation methods are needed to estimate the blurred length and blurred
angle. The blurred angle can be estimated by using the Gabor filter to filter the blurred
image in the frequency domain, and the blurred angle can be estimated according to
its response. The fuzzy length can be used to train the neural network to estimate the
ancient fuzzy length. After obtaining the estimated values of blurred angle and blurred
length, the fuzzy point spread function can be constructed. Finally, Wiener filtering
can be used to restore the blurred image according to the point spread function.
4.2.2 Estimation of Image Blurred Angle Based on Gabor
Filter
(1) Gabor filter basics
The motion blurred image can be found according to the spectrogram to show that
the motion blurred image has obvious directional characteristics in the frequency
domain.