Applied Mechanics and Materials Vols. 411-414

Abstract: A novel watermarking technique to authenticate video of H.264 is presented in this paper, using Uniform Content Locator (UCL) to semantically indexing video content and dual watermarks to preserve and enhance video content integrity and authentication. UCL index information is firstly extracted from video content and is formatted as semantic watermark to be embedded in video content. The UCL watermark is regarded as robust watermark and is then embedded into medium frequencies of DCT-coefficients of H.264 video I-frames in order to protect video attributes property (e.g., video author, copyright, content category). Features information obtained from the previously watermarked DCT-coefficients are treated as fragile watermark and are embedded into the motion vectors of H.264 video P-frames in order to ensure video secrecies (e.g., video integrity, authentication). Experiments demonstrate that this proposed technique can fulfill the requirements of H.264 video authentication and has negligible effects on video code rate change and content distortion.

Abstract: In content-based image feature extraction research areas, SIFT feature occupies a very important position. In 2004 it was first proposed, widely used in object recognition, video tracking, scene recognition, image retrieval and other issues, and achieved great success. But the extraction of image SIFT features needs a huge amount of computation. This paper presents the concept of color energy in where it has great information, and extract large color energy regions, extracted SIFT feature points in them. Although losing some of the feature points, this method effectively reduces the computational complexity, and reduces the computation time.

Abstract: It is stated in this thesis that image compression is realized by wavelet transform and the advantages of wavelet transform are pointed out. Based on the evaluation criteria of the performance of image compression, a comparison is conducted with the traditional DCT transform compressed image. It is indicated by the result that the performance of image compression based on wavelet transform is better than that of based on DCT transform.

Abstract: This paper proposes a digital watermarking method for the protection of medical volume data based on chaotic neural network and 3D DWT-DFT. In order to improve the security of digital watermarking, it uses chaotic sequence for scrambling .The chaotic sequence is generated by chaotic neural network. It makes full use of multi-level decomposition characteristics of 3D DWT-DFT, embedding watermark in the transform domain. The experimental results show that the method can resist filtering and rotation attacks, which has good robustness.

Abstract: The pinhole model of camera imaging is established based on the analysis of the geometrical model of camera imaging and the mapping relationship and calculation formula between target and image are deduced in ideal status. Then the model is improved to make the model more practical, and a lens first-order radial distortion of the pinhole camera model is established. Final, the stability of the model is effectively analyzed.

Abstract: A method based on Gabor filter bank is presented to automatically detect the steel cord conveyor belt fault. Firstly, multi-channel filtering is implemented by Gabor filters. Then the characteristic differential and fusion images are calculated. Finally, the fault images are identified by white pixels statistics after the threshold segmentation. The results show that this method has good detection result for steel cord conveyor belt fault.

Abstract: A novel algorithm based on gradient image histogram threshold and lowest mean edge value is proposed. Image objects are initialized using the gradient image histogram threshold and grow up using the watershed segmentation algorithm. To reduce the number of image objects, a merge method based on lowest mean edge value is proposed. The segmentation result revealed the advantage of this method in preventing over-segmentation.

Abstract: In this paper, facial expression recognition as the starting point, it is extracted the mixing characteristics of the facial expression, including geometric and texture features. It can solve the problem of the collection contains the degree based on Variable Precision Fuzzy Rough Set, It could improved the accuracy of facial expression recognition using the method, thus making the facial expression recognition process more accurate and efficient.

Abstract: Aiming at the requirements of medical image registration for good robustness, high-accuracy and speed, this paper proposes a Medical Image Registration algorithm Combined SURF with improved RANSAC algorithm. This algorithm first of all extracts SURF featured points from images and matches similar featured points, then the improved RANSAC algorithm is used to eliminate wrong matches, and finally the image registration process is accomplished by estimating space geometric varied parameters according to least square method. The algorithm combines robustness and high efficiency of SURF and high-accuracy of improved RANSAC algorithm. Experimental results show that in the condition of images with noise, non-uniform intensity and large scope of the initial misalignment, the proposed algorithm achieves better robustness and higher speed while maintaining good registration accuracy compared with the conventional ional area-based and feature-based registration methods.

Abstract: This paper presents a Thai character recognition method based on topological properties. The method first extracts gradient features from a character image. A two-step classification are then applied to recognize the character. In the first step, a conditional random fields model is used to generate a set of possible characters. Then a nearest neighbor model based on hierarchical centroid distance is employed to finally recognize the character. The proposed method is trained by printed characters from documents and vehicle license plates. The technique is evaluated and found to have the recognition rate of 96.96%.