To solve the copyright protection of stereo images, a robust stereo image watermarking method based on correlations of left and right views is proposed. Because the tucker decomposition can preserve the main energy of the image well, tucker decomposition is performed on left and right views, respectively, to make full use of the correlations of three channels in the color view. Each view is decomposed into three feature images, where the first feature map retains relationships of three channels in each view. Secondly, considering correlations between the left and right views, the first feature images of the left and right viewpoints are combined to be performed by using tucker decomposition, and the main energy features images of the stereo image are obtained. Finally, the main energy feature image is decomposed by singular value decomposition, and watermark is embedded for the purpose of improving the robustness. Experimental results show that the proposed method has high robustness and invisibility, and realizes blind extraction of watermark.

In the process of collecting hand vein images, due to the influence of image acquisition equipment, illumination and subcutaneous fat thickness, the contrast of hand vein images is relatively low. Meanwhile, vein extraction is seriously affected by image noise. To solve this problem, an algorithm of image segmentation and contrast enhancement based on features of venous gray value is proposed in this paper. Firstly, effective region of interest (ROI) is extracted and filtered through Wiener filtering. Secondly, a new image segmentation algorithm is obtained to extract vein image. The venous binary image is denoised by an 8-adjacent inner boundary tracking method and morphological processing. Finally, contrast-enhanced venous images are obtained by weight stack of the ROI and denoised images. The experiments results show that intravenous veins can be obtained perfectly by using the image segmentation algorithm based on features of venous gray value. Moreover, the high contrast venous images can be obtained.

This study adapted simulations to analyze the fitting capabilities to human aberration of three kinds of Bimorph Deformable Mirrors (DMs) with different spatial resolutions, especially the capability to fit to 3~ 35th orders of Zernike static aberrations and the human aberrations, including the eye of diseases. It’s shown that Bimorph DM is well suitable for fitting to low-order aberrations with the error less than 0.15. As the spatial resolution increased, the capability of fitting aberrations enhanced totally. Compared with traditional discrete piezoelectric DM, 35-element Bimorph DM had smaller fitting error on the first 20th Zernike aberration. This simulated analysis provided an analytical method for the selection of Bimorph DMs for high-resolution human eye imaging systems. In addition, it provided a research foundation of further improvement of the Bimorph DM in fitting aberrations capability.

The current dual-mode infrared image lacks the selection and combination basis of each element when constructing the fusion method, and the fusion model cannot dynamically adjust for the image difference feature, resulting in poor fusion effect. Aiming at the above problems, referring to the multi-parameters of biological characters, this paper proposes infrared intensity and polarization image mimicry fusion based on the combination of variable elements and matrix theory. Firstly, the fusion model was divided into four parts: fusion algorithm, fusion rule, fusion parameter and fusion structure. The single mapping relationship between different parts and the difference of image feature fusion is established. Secondly, using the imaginary transformation idea, the imaginary transformation fusion method was established, and the necessary four parts of the fusion process are combined to derive a new fusion algorithm. Finally, it used the different source images with different features to verify the proposed mimetic fusion algorithm. Experimental results show that when the image difference features are different, the fusion method was more suitable for deriving image features, so as to achieve active selection and adjustment of the fusion algorithm. The different features in the fusion image can be effectively combined, and the visual effect of the original image is significantly improved.

In order to solve the problem of imaging drift in scanning electron microscope (SEM) that caused by electron beam drift, electromagnetic interference and other reasons, an image shift correction algorithm based on ORB (oriented FAST and rotated BRIEF) combing the PROSAC (progressive sample consensus) is proposed in this paper. Firstly, the ORB algorithm is used to detect the feature between the reference image and real-time image. Then the initial matching of the feature is implemented by using the Hamming distance and cross-matching. Moreover, the RANSAN (random sample consensus) optimization algorithm PROSAC is used to calculate the homography matrix between frames and the final exact homography matrix is re-iterated after eliminating exterior point. Finally, the SEM image drift is corrected in real time using the perspective transformation of the homography matrix. The experiments show that the proposed algorithm is high precision and satisfies the requirement of SEM real-time processing.

In order to build a robust background model and improve the accuracy of detection of foreground objects, the temporal correlation of pixels at the same position of the video image and the spatial correlation of neighboring pixels are considered comprehensively. This paper proposed a background modeling method based on multi-feature fusion. By using the domain correlation of pixels in a single frame image to quickly establish an initial background model whichis updated using pixel values, frequency, update time and sensitivity of the video image sequence, the ghost phenomenon is effectively improved and the holes and false prospects for moving targets are reduced. Through multiple sets of data tests, it shows that the algorithm improves the adaptability and robustness of dynamic background and complex background.

In order to comprehensively and objectively study the degenerate factors of underwater turbulent imaging and optimize the corresponding image restoration algorithms, a reusable submarine imaging experiment system with a controllable turbulent flow condition is established. The circulating water pump is used to control the intensity of turbulence in the laboratory tank.The bubble generator is used to generate micro bubbles. The image sensor is used to obtain the images of sinusoidal stripe target plates under different conditions. The effect of turbulent flow field, path radiation and fluid media on submarine imaging in turbulent flow were studied, and the differences and applicability of modulation transfer functions (MTFs) of three degradation factors are compared by combining image restoration and super-resolution reconstruction. The experimental results show that the turbulent flow field causes MTF declines of the low spatial frequency, and the path radiation and fluid media lead to the decrease of modulation contrast of the high spatial frequency. In the restoration of the underwater turbulent degraded image, the MTF of the turbulent flow field is suitable for image restoration, and the MTFs of the path radiation and fluid media are suitable for image reconstruction.

The selection of the focusing window is the key procedure in achieving precise automatic focus of the microscope. For the traditional focus window selection method, the limitation is that the target object cannot be accurately positioned. This paper proposes an improved artificial fish focusing window method. The method takes the area-of-interest of the whole image as the basis of the selection window. Through utilizing the global optimization ability of the artificial fish swarm algorithm, the best focusing window can be obtained. Adding the global optimal value to the behavior update of each artificial fish makes the artificial fish quickly move to the optimal position. Under the premise of ensuring accuracy, the elimination behavior is introduced to improve the convergence speed of the algorithm in the later period. Furthermore, according to the characteristics of the bulletin board in the algorithm, the interference area is identified with the trend comparison method, and the influence of the non-target area is effectively excluded. Experiment results show that the focusing window obtained by this algorithm can be well-suited for the target object, greatly improve the accuracy of autofocus, and make the efficiency improvement 1.65 times than the traditional method.

Aiming at the problem that 3D LiDAR point cloud has high data density, outlier noise, and scattered distribution in urban environment, which is not conducive to the matching between point clouds in the later stage, a pre-processing method for large-scale LiDAR point cloud frame matching in urban environments is proposed. First, the point cloud data is transformed into a Mean Elevation Map, and the ground point segmentation processing is performed on the point cloud using the height gradient between the grids; then, the DBSCAN clustering algorithm is improved by the three-dimensional voxel grid division method, and the improved VG-DBSCAN is used to cluster point clouds and separate the target point cloud from the outliers after clustering, thereby, which eliminates outlier noises in the point cloud. Finally, the Voxel Grid filter is used to down sample the point cloud. The experimental results show that the proposed method can perform real-time preprocessing on point cloud data, and the average time is 132.1 ms. After pre-processing, the accuracy of point cloud frame matching is increased by 2 times, and the average time consumption is only 1/6 before pre-processing.

Focusing on the airplanes in remote-sensing images, a real-time algorithm based on improved YOLOv3 is proposed to detect airplanes in remote-sensing images. Firstly, a convolutional neural network that consists of 49 convolutional layers is proposed to detect airplanes in remote-sensing images specifically. Secondly, dense connection is employed on proposed convolutional neural network, and maxpool is employed to enhance the feature transmit between dense blocks. Finally, to deal with the fact that airplanes in remote-sensing images are small targets mainly, we propose to increase the scale detection from 3 to 4 and employ dense connection to merge feature map among different scales. The algorithm is trained and tested on the designed airplane dataset. The experiment results show that our algorithm obtain 96.26% on precision and 93.81% on recall.