Credit risk management is an integral part of banking business. By modifying the standard Markowitz model, the paper aims to study optimum credit portfolio with binary variable, which decides whether providing loads. The model can evaluate the cumulative risk and returns by considering the asset correlation and whether providing loads in the prospective of influencing risk indicators and profit rates. It enforces the empirical research of optimum portfolio by using the proposed model in this paper and the statistics offered by Standard & Poor's rating agency.

The main contribution is to present stability analysis criterion for a type of norm⁃bounded uncertain perturbed discrete systems(NUPDS) with its states is subjected to a novel ellipsoidal outer approximation,which is different from the traditional bound description of general state. Based on the Lyapunov method, a delay dependent criterion and the corresponding state feedback control law, which are expressed in terms of linear matrix inequalities(LMIs) technology, are proposed to guarantee the asymptotic stability for the perturb delay systems. The solution of the LMIs can be obtained easily using existing efficient convex optimization techniques. A numerical example is given to illustrate effectiveness of the proposed method.

Reconstruction of rippedup documents was important for judicial authentication，cultural relic reconstruction and military information acquisition，etc．An approach for the reconstruction of regularly rippedup documents was proposed. Firstly, the line spacing of the fragments was automatically extracted. Secondly, meanshift algorithm was adapted to cluster all the fragments. Thirdly, for each cluster, the similarity matrix was constructed based on the grayscale of the edges of the fragments. Lastly, the document with genetic algorithm was reconstructed.Reconstruction of rippedup documents is important for judicial authentication，cultural relic reconstruction and military information acquisition，etc．In this paper，we propose an approach for the reconstruction of regularly rippedup documents. Firstly, the line spacing of the fragments is automatically extracted|Then meanshift algorithm is adapted to cluster all the fragments|Then for each cluster, the similarity matrix is constructed based on the grayscale of the edges of the fragments|Lastly, we reconstruct the document with genetic algorithm.

Embedding and extracting algorithm for watermarking in the frequency domain based on the Fresnel and Fourier Hologram was proposed. The image watermarking was restored from the JPEG images with different compression ratios and the stability of watermarking image was compared between two different transformations. The result show that the image watermarking obtained by the proposed algorithm had perfect imperceptibility and stability. By comparison, however, the one based on the Fresnel Hologram had superiority in terms of stability.