Abstract Image inpainting refers to the process of filling in missing regions or removing objects, and has broad application prospects. The rapid development deep learning led new technological breakthroughs image repair technology, continuously improving quality inpainting. However, when we inpaint large regions, texture structural features cannot be comprehensively utilized. This leads blurry images. To solve this problem, propose an improved dual-stream U-Net algorithm that adds attention...

With comprehensive consideration of generalized synchronization, combination synchronization and adaptive control, this paper investigates a novel complex (AGCCS) scheme for different real nonlinear systems with unknown parameters. On the basis Lyapunov stability theory an AGCCS controller parameter update laws are derived to achieve identification two drive response system, as well system. Two simulation examples, namely, ACGCS chaotic Lorenz Chen driving hyperchaotic Lü presented verify...

The fingerprint segmentation algorithm based on gray variance can't segment those images with high noise. After analyzing limitation of the image variance, paper proposes improved to acquire average and combining basic distributing character in valid region. Experimental results indicate that are more representative original information region than classical algorithm. segmented proposed exact reliable.

Fingerprint classification is crucial to reduce the processing time in a large-scale database. In this paper fingerprint based on continuous orientation field and singular points proposed. The can not only filter noises point directional image, but also represent basic structural feature of more precisely. Singularities are most important reliable classification. fast algorithm made possible by simple effective combination modified Poincare index determination points.The experiment results...

Effective fingerprint classification not only can provide an important index mechanism for large database, but also improve the efficiency and performance of AFIS. At present, because traditional Poincare method detection more false singular points weaker anti-noise problem, this paper presents a based on continuously directional image symmetrical axis. Compared with algorithms, algorithm has following two aspects improved: firstly, exhibits good continuity, well gradualness, excellent...

To achieve higher education internationalization and cultivate the more new inter-disciplinary talents, it is necessary to implement bilingual teaching in computer speciality. How further improve push work a level from its current situation currently an important subject be discussed reformation of teaching. A perfect effective method based on years experience practice proposed this paper. In addition, necessity implementation existing questions speciality discussed.

To interpolate mine pressure missing data, it is proposed that cubic spline interpolation algorithm based particle swarm optimization. The use optimization to solve the coefficients of function, in order reduce tedious solving interpolation. By analysis data different working face same place, and difference interpolated results actual values are close, therefore, effective on for data.

The classification of Thangka headdress has a wide application in the semantic retrieval and annotation Thangka. existing methods are facing difficulties segmentation, which affects practical application. In order to improve efficiency, this paper proposes new method using support vector machine (SVM) for with combination multi-features. adopts following steps: firstly, segment Kirsh segmentation method, secondly, extract features Hu moments, Fourier Zernike thirdly, color features, total...

Data Structure is the core course in computer specialty, which aims to cultivate undergraduates' abilities of selecting appropriate data structures develop high quality application programs. CDIO teaching model has already become one hotspots and effective solutions improve effect. By learning advanced engineering education analyzing problem traditional teaching, we try apply Structure. Taking projects as a vehicle, article puts forward some concrete reform measures from curriculum system,...

Damaged regions must be segmented before digital repairing Thangka cultural relics. A new segmentation algorithm based on local distinction is proposed for segmenting damaged regions, taking into account some of the area with a transition zone feature, as well difference between and their surrounding combining gray value, complexity definition-complexity (LDC). Firstly, calculate normalized; secondly, thirdly, distinction; finally, set threshold to segment image, remove over segmentation,...

As a representative of Tibetan culture, people's headdresses in Thangka can be divided into hairpin, monk hat and crown. In order to meet users' demand for accurate retrieval Thangka, the category information used mark thereby increasing accuracy. Existing headdress classifiers suffer from common problem: image segmentation is required before classification. When not satisfactory, human interaction also required. This paper presents classification method based on convolutional deep neural...

Image inpainting is to repair target regions such damaged or an object that you want remove. Exemplar-based technique cannot find suitable similar patch sometimes when matching using only color distance. Our work improve exemplar-based technique. The main includes: (1) adding boundary-constrained patches more patches; (2) boundary filling instead of filling, reducing the possibility misfiling; (3) searching for in four rotation directions patches. experimental results show proposed method...

Line segmentation is an important step in image character recognition. However, due to the problems of interline adhesion, overlapping, and skewing document images, effect text line not ideal. Therefore, further research on Tibetan ancient books urgently needed. The standard A* algorithm ideal for books, so this paper proposes a block-based books. method firstly preprocesses such as binarization tilt correction then performs horizontal projection. obtained projection histogram subjected...

In the teaching of software engineering for computer related majors, there is a gap between high-level logic design and implementation, which leads to students' unreasonable design, implementation code does not follow scheme. Therefore, it necessary carry out heuristic students, starting from programming students are most familiar with, gradually transition using reverse thinking mode, so that can master transformation relationship implementation. case teaching, restore or demand analysis...

In order to reduce the computational cost in current MIMO-ISAR back projection imaging algorithm, a fast algorithm based on FFT is proposed for projection.First, we establish reflected wave signal model imaging.The approximate representation of linear change phase compensation factor provided.Second, constraints conversion from non-linear given.Finally, an efficient achieved.Simulation results show that compared with traditional methods, reduces greatly.

Data Structure is the core course in computer specialty, which aims to cultivate undergraduates' abilities of selecting appropriate data structures develop high quality application programs. Research teaching has already become one hotspots and effective solutions improve effect. In this paper we explore principles practices based on research teaching, including curriculum system, building a multi-level system practical course, resources platforms, methods, bilingual team cooperation...