As a new non-invasive and high-resolution scanning method, optical coherence tomography(OCT) has been widely used in clinical practice, but OCT images haveserious speckle noise, which greatlyaffects the diagno-sis of diseases. Two original dictionary noise reduction algorithms have been improved for multiplicative speckle noise in OCT. The algorithm first performs logarithmic transformation on OCT images, uses orthogonal matching pursuit algorithm for sparse coding, and K singular value decomposition learning algorithm to update adaptive dic-tionary. Finally, it returns to the space domain through weighted average and exponential transformation. The expe-rimental results show that the improved two dictionary algorithms can effectively reduce the speckle noise in OCT images and obtain good visual effects. The noise reduction effect is evaluated by four factors: mean squared error (MSE), peak signal-to-noise ratio (PSNR), structural similarity (SSIM) and edge-preserving index (EPI). Compared with the two original dictionary noise reduction algorithms and the traditional filtering algorithms, the noise reduction effect of the two improved dictionary algorithms is better than that of other algorithms, and the improved adaptive dictionary algorithm performs better.

Multi-frame image super-resolution method fuses the information of multi-frame low-resolution images to reconstruct high-resolution images. For multi-frame image super-resolution, the accurate estimation of blur kernel of low-resolution image is prerequisite for efficiency information fusion. Traditional super-resolution method usually assumes a known blur kernel and uses the Gaussian filter blur kernel for the enhancement. It also needs to tune the parameters by time-consuming hand-tuning. The proposed method acquires the super-resolution method based on the variational Bayesian method. The high-resolution image, the blur kernel and the model parameters are estimated simultaneously and automatically in the optimal stochastic sense. Experiments and simulations demonstrate that the proposed blind super-resolution method based on blur kernel self-adaptive estimation outperforms the state-of-art super-resolution method in variational Bayesian framework, especially, for the high signal to noise ratio scenarios.

The short-distance free-space transmission characteristics of the optical orbital angular momentum (OAM) beam were experimentally studied. The transmission distance is 0～50 m indoors. A digital micromirror device (DMD) was used in the experimental setup to generate the OAM beam. At the receiver, a spatial beam analyzer was used to measure the intensity pattern of the OAM beam. The beam broadening effect of the OAM beam at different trans- mission distances was studied. The phase pattern of the OAM beam was studied by the interferometric method. At the receiver, a single path Sagnac interferometer (SPSI) was used to separate and detect the intensity of modes of the OAM beam. The effects of energy migration from the sending mode to the sideband modes of the OAM beam were studied. A heater was used to generate the strong turbulence to simulate the influence to the OAM beam mode transmission characteristics.The experimental results show that the OAM beam with large mode topological charge has more deterioration of the mode purity after transmission in the strong turbulence.

Existing fine-grained categorization models require extra manual annotation in addition to the image cat-egory labels. To solve this problem, we propose a novel deep transfer learning model, which transfers the learned representations from large-scale labelled fine-grained datasets to micro fine-grained datasets. Firstly, we introduce a cohesion domain to measure the degree of correlation between source domain and target domain. Secondly, select the transferrable feature that are suitable for the target domain based on the correlation. Finally, we make most of perspective-class labels for auxiliary learning, and learn all the attributes through joint learning to extract more fea-ture representations. The experiments show that our model not only achieves high categorization accuracy but also economizes training time effectively, it also verifies the conclusion that the inter-domain feature transition can acce-lerate learning and optimization.

The calculation of correlation is critically important for ultrasound strain elastography. The sum-table based method for the calculation of the normalized correlation coefficient (ST-NCC) can greatly improve computational efficiency under an environment of serial computing. Its implementation and performance are yet to be investigated when given a parallel computing platform, particularly, under a GPU environment. In this study, the published ST-NCC method was implemented into GPU and its performance was evaluated for speckle tracking. Particularly, the performance of the ST-NCC method was compared to the classic method of computing NCC using simulated ultrasound data. Our preliminary results indicated that, under the GPU platform, the implemented ST-NCC method did not further improve the computational efficiency, as compared to the classic NCC method implemented into the same GPU platform.

Abstract: Interferometric hyperspectral image is a special kind of image source, which contains massive data and is difficult to transmit on a limited bandwidth channel. The traditional method is to compress the data and then encode the transmission. However, the compressed data is still very large, which brings great difficulties to the transmission and storage of data. Nevertheless, the compressed sensing technology can solve this problem well. Based on the original algorithm of compressed sensing, this paper proposes an adaptive threshold-based orthogonal matching pursuit algorithm (ATROMP) which is more suitable for interfering hyperspectral images. The algorithm first uses block processing and then selects the interference fringes. Because the vertical interference fringes have strong unidirectional characteristics, the total variation of the level is larger.Therefore, the interference fringes in the images are extracted from the horizontal total variation values for adaptive sampling. Then, an adaptive threshold is used in this paper to replace the quadratic selection in the ROMP algorithm. Using an adaptive threshold can not only ensure that the atomicity of each selected atom is sufficiently high, but also that multiple atoms can be properly selected each time to ensure sufficient number of cycles, to avoid the follow-up higher degree of atom missing. Compared with the traditional ROMP algorithm, a large amount of experimental data show that the sparse reconstruction ac- curacy of the method can be significantly improved.

The photocathode of solar-blind photomultiplier tube is Te-Cs, and the glass tube is synthetic quartz. We use special oven with two lights to prepare solar-blind photomultiplier tube. We test the performance of solar-blind photomultiplier tubes. Results show that CR340 has good solar-blind characteristic, cut-off wavelength at 320 nm, and output sensitivity can reach 5×105 A/W (250 nm), and gain can reach 1.3×107, and life is more than 1000 h. Solar-blind photomultiplier tubes are evaluated by several domestic analytical instruments manufacturers, all feed-back are very good.

Aimed at the problem of automatic focus and image system quality evaluation in microscopy imaging, a micro-image definition evaluation method is presented by combining multi-scale decomposition tools and absolute gradient operators. The multiscale and multidirectional non-subsampled Shearlet transform is utilized to decompose the input micro image into a low frequency sub-band image and a number of high frequency sub-band images. Combined with the anti-noise threshold setting, the gradient absolute sum values of each sub-band image were calculated. By using the different effects of image sharpness on the low-frequency and high-frequency sub-band coefficients, the ratio of the high-frequency to low-frequency gradient absolute value operator was taken as the final evaluation value of the microscopic image sharpness. The simulation experiment and actual experiments were car-ried out and the experimental results illustrated that the proposed approach has good monotonicity and anti-noise characteristics. Compared with other classic evaluation algorithms, the presented method obtained superior per-formance on sensitivity, stability and robustness. It has very good practical application values.

In order to overcome the problem of over-segmentation caused by traditional watershed algorithm, a colorimage segmentation algorithm based on simple linear iterative clustering (SLIC) and watershed algorithm is pro-posed to achieve an ideal segmentation effect. Firstly, the algorithm calculates the number of super-pixelspre-segmented by image complexity, and uses SLIC to super-pixel segmentation preprocessing of the original imageto reduce the redundant information in subsequent processing. Then, an adaptive threshold calculation method isproposed to process the gradient image of the preprocessed image in order to effectively remove noise and obtainmore complete contour information. Finally, the watershed segmentation algorithm is used to segment the imageextracted by the minimum value marker. Experiments on a large number of images show that the proposed algorithm can effectively suppress the over-segmentation problem caused by the traditional watershed algorithm, and is su-perior to the traditional algorithm in the comparison of LCE and GCE, and the segmentation quality is improved.

In order to achieve electrowetting real-time display, a display driving system, consisting of a DVI video codec system and FPGA timing control system, is designed. DVI video codec system is responsible for obtaining signal sources and for image coding and decoding. FPGA is responsible for buffering and processing of video data and for controlling electrowetting driving waveforms. This paper also proposes an improved multi-grayscales dy-namic symmetrical driving waveform, which improves the ink-splitting phenomenon and suppresses the charge-trapping phenomenon while increasing the gray level. The results show that the driving system successfully improves the problems of oil-splitting and charge-trapping, and drives the 1024×768 resolution electrowetting display to play video in real time following the computer. The frame rate of the video reaches 60 frames/second, and the highest gray level of the pixel reaches 15. These properties meet the requirements for dynamic display of the electrowetting paper.