Restoration Of Blurred Images Of Standing Trees In Case Of Random Motion

When robot is walking on the soft or moist forest floor on the complex forestry region, it is prone for it to slip and slide. The slip and slide without any law could cause motion blurred images to the forestry robot vision system. Thus it affects the accuracy of working environment recognition and environment modeling. When some large forestry robots are working, irregular trembles and vibrates often occur, which generates the motion blur to the images from vision system of forestry robot. So it is necessary to research on the restoration methods about motion blurred images of standing trees in the case of random motion for the forestry robot.Domestic and foreign scholars study extensively on the motion blurred image restoration methods at present, but motion blurred image restoration problem in special natural environment is rarely reported to the forestry robot. The restoration method of motion blurred image of standing trees to forestry robot with random motion is researched in this paper. The main contents and innovation as follows:1. A de-noising algorithm based on robust joint sparse coding (RSSC) is proposed for the de-noising problem to motion blurred image of standing trees. Image block similarity is effectively utilized by joint sparse regularization, the difference between the image block is in the view of elements sparse regularization, which could improve the effect of de-noising. Comparing with K-SVD and BM3D algorithm, the experiment shows that the SSIM value of RSSC algorithm is higher than K-SVD and BM3D algorithm with different Gaussian noise (?)n=15,25,35.2. In view of the defects of long time calculating and large computation for motion blurred angle identification in restoration method of uniform linear motion blurred image, an automatic identification method of improved motion blurred angle is proposed. Making segmentation of the target image and changing the step as descending to identify the motion blur angle, removing the angle with large differences, then the motion blurred angle will be get after taking the average value. By comparing with the traditional method in the experiment, the method in this paper can reduce the calculation time and improve the computational efficiency, its accuracy is better than traditional methods.3. A sparse regularization motion blurred image restoration method based on dictionary migration is proposed in this paper. This method take the image block collected from the image that need to be restored as a training sample, and it takes the global trained dictionary to blurred images which are special and need to de-motion, this method studies a set of adaptive learning dictionary while improving the accuracy. Comparison is adopted among motion blur kernel, Gaussian blur kernel, uniform blur kernel and Krishnan&Dilip algorithms in the experiment, the SSIM value in this paper is better than the other two algorithms, the restored image is much more close to the original image and it solves the ringing effects.4. Motion image restoration method based on wavelet frame is studied and a new mathematical model is proposed. In the new model, two regularization items are introduced by given initial blur kernel; one is to reduce the large number of wavelet frame coefficients which can be seen as reducing the number of discrete pixels. The other is to compensate the error caused by‖Wp‖1through controlling the remaining minimizer to deviate to the larger connection support. By adjusting the balance of the two regularization terms, the method gets the reasonable motion blur kernel and restores the original image with Bregman method. Blur image experiments are taken with uniform linear motion, uniform curve motion and erratic motion, comparing the data with the methods of Shan et al. and Fergus et al., the restored effect for irregular motion blur image in this paper is the best among the three methods, and the SSIM is the highest. The noise immunity is analyzed in this paper. The results show that the method in this paper has a strong resistance to salt and pepper noise.