Recent Patents on Computer Science - Volume 10, Issue 1, 2017

Background: In the field of pattern recognition, there is a big problem for mobile image feature recognition technology. To solve these problems is one of the hot spots in the present study. Method: This paper analyzes the current developments and recent patents in the mobile image feature recognition. A variety of different methods and patents based on color recognition, texture recognition and shape feature recognition are introduced. Various identification methods' principle and advantages and disadvantages are made, detailed explanation, and the comparison results of various implement methods are given. Result: Improved algorithm can improve the Identification number, recognition speed and recognition accuracy. Conclusion: Improved algorithm can provide theoretical reference to further solve the characteristic and difficulties in the mobile image feature recognition.

Background: Multimedia information processing technology is an important part in the field of computer information processing, as described in various patents. Method: In order to solve the problems existing in different methods in this field, this paper focused on image processing, audio processing, video processing, 3D animation technology and other information processing technology and the application of the technology's frontier approach to make a detailed theoretical analysis and describe their application. The principle and the prospect of the application and the existing problems were also analyzed. The validity of the proposed viewpoint was proved by experiments. Result: Improved algorithm significantly increased the quantity of processed information and processing efficiency, and reduced the error rate. Conclusion: Through the summary of multimedia information processing technology, an effective basis for the related research is provided.

Background: In the increasingly large scale of the current multimedia information database, as described in various patents, big data technology is becoming more and more mature, and the problem of massive multimedia data mining is becoming more and more important. Method: In order to solve these problems through the analysis of the status for multimedia data mining, the general system structure in the application of multimedia information mining technology is analyzed, and the different methods and principles of text mining, image mining, video mining, web data mining in multimedia information data mining are made detailed analysis and the advantage and disadvantages are contrasted. Through the design of different experiments, the effectiveness and inferior of different methods are verified. Result: In the improved algorithm, the mining result of the multimedia text data based on correlation analysis is very different from the actual multimedia text data, the result is closer to the original mining image result, so this method is more practical. Conclusion: Through the summary and discussion on the problems existed in the current method, the faced challenges of multimedia information data mining technology are pointed out, to lay the foundation for the further development of the related technology.

Background: With the development of three dimensional (3D) biometric recognition technologies, the 3D image recognition technology has encountered a great development space, and existed larger development bottleneck, also described in various patents. For example, in the process of imaging, there must be a lack of three-dimensional information of objects, resulting in low recognition rate. Method: This paper analyzes the progress and technology development process of 3D biometric image recognition and discusses the main difficulties and future development direction. In these methods, the minimum distance method, correlation calculation method, nearest vector method, principal component analysis method, Markov model, support vector machine method and other various methods of 3D morphology are introduced in detail. Result: The average recognition rate of the harmonic analysis algorithm is 98.5%, higher than the recognition rate of the ICP algorithm, which is 96.4%. Conclusion: The experimental results verify the effect and the advantages and disadvantages of different methods, to provide the relevant content for studying on recognition of the 3D biometric features.

Background: Three dimensional image feature reconstruction is to establish a mathematical model, which is suitable for computer representation and processing. It is the basis of processing, operating and analyzing its properties in computer environment, as described in various patents. Method: This paper introduces the present situation of the research on the reconstruction of 3D image features, and analyzes the method of 3D image reconstruction. 3D feature image reconstruction is an important branch of 3D image processing technology. Through the principle analysis and technology research of 3D image reconstruction technology, the development status of the relevant research at home and abroad is made detailed analysis. The different methods for surface rending technology, volume rending technology, and stereo vision matching technology are analyzed; in addition, the working principles and the existing problems are analyzed. Result: Experiments have proved the performance of the correlation method, e volume rendering technology can remodel real three-dimensional shape of the corn leaf better, the ups and downs of edge of the corn leaf and drape, other details can be preserved, effectively improve the convergence speed, and greatly reduce the running time and the number of iterations. Conclusion: It provides the development direction for the 3D image feature reconstruction.

Background: In order to further analyze the problems of remote sensing image processing technology and the future development direction, this paper analyzes the concept of remote sensing and basic process, describes the status of domestic and international development of remote sensing image processing technology, as described in various patents. Method: Several key elements of remote sensing image processing technology are studied, including five directions, remote sensing image matching, the integration of infrared and visible light remote sensing image, edge detection of SAR images, remote sensing image classification and remote sensing image change detection, and research and development status of these five directions were analyzed. The main characteristics of remote sensing image processing technology and reasons constraining the development of remote sensing image processing technology are analyzed. Result: Image fusion method based on PCA, are relatively vague and inaccurate. The image fusion method based on wavelet transform is used for image fusion, the fusion result is clear, and the image fusion method based on wavelet transform is more accurate than PCA. Conclusion: The comparison results of different methods are given by computer simulation experiment; it provides a reference for the development trends and application prospects of remote sensing image processing technology.

Background: Aiming at the problems and defects of different methods in current multimedia image segmentation. Method: The more advanced multimedia image segmentation techniques are introduced in this paper. The threshold segmentation, edge detection, region extraction technology have made detailed analysis and introduction. And in view of the image segmentation techniques under specific algorithm, these methods are introduced, meanwhile, the principle of these methods and the existing problems, as described in patents, are analyzed. Result: In the improved algorithm, the small region after segmentation is obviously reduced, the time required for region segmentation method is shorter, the accuracy is higher than the threshold segmentation method. Conclusion: The simulation results show the advantages and disadvantages of the different algorithms, which can provide reference for further development of multimedia image segmentation technology.

Background: Image segmentation is a very important problem in the field of image processing, complete and accurate image segmentation technology in many fields have extremely important applications, as described in various patents, but current segmentation methods in different applications exist great defects. Method: This paper makes a detailed analysis of the development trend of image segmentation technology. And the methods based on the boundary segmentation, threshold segmentation, region growing segmentation, random field segmentation, neural network and wavelet calculation are introduced in detail, and the application direction and existing problems are summarized. Results: It can be seen that the image segmentation algorithm based on wavelet transformation has a better segmentation effect, Reasonable division of different types of images. Conclusion: Experiments have proved the application property of different methods, providing theoretical guidance for the design of a real-time, robust algorithm.

Background: Scene image classification is a fundamental problem in the field of computer vision, as described in various patents. But, so far, it is still a challenging task to solve the semantic gap of the scene image between low level feature and high level topic. Method: In this paper, we propose a new scene image classification method based on super pixel segmentation and correlated topic model. The method is composed of the following steps: Firstly, considering super pixel providing the spatial support for computing region, we divide image into sub-regions through super pixel segmentation model. Then, each sub-region is described by lots of local scale invariant feature transform key points. In order to preserve the mode information of key points, we use Median- shift method to build word of bag to represent image. Lastly, in order to reflect the relation of the low level features and the high topics of images, we use a correlated topic model based on word of bag to classify scene image. Result: We evaluated the proposed method on the classical Caltech 10 database. The experiment results show that the presented method have average precision rate with 72.6% for scene image classification. Conclusion: From the experimental results we can draw the conclusion that the super pixel segmentation method can preserve more spatial support to scene image, and the correlated topic model can mine the high-level semantic information scene categories from low-level feature, which make the presented method highly completive than other approaches.

Background: Image fusion is a fundamental issue in image processing and data fusion, as described in various patents. Its purpose is to synthesize important scene information from two or more images to a fusion one, which is more adaptive to human vision system or sub-sequential processing, such as image classification, object tracking and so on. Method: In this paper, we proposed a novel image fusion method based on Shearlet Transformation, which is thought to be a suitable multi-scale geometric analysis tool to represent image. Firstly, we decomposed the source images into the shearlet domain. Secondly, we designed two kinds of measurements with local information in the same subband and the cousin coefficient in other direction subband respectively. Result: Then, we make a fusion criterion by incorporating the two measurements mentioned above. Last, the fusion image is obtained by inverse Shearlet Transfomation. Conclusion: The experimental results demonstrated that our proposed method outperformed the compared state-of-the-art fusion algorithms.

Background: Deep neural network based methods have obtained great progress in a variety of computer vision tasks, as described in various patents. But, so far, it is still a challenge task to model temporal dependencies in the tasks of recognizing object movement from videos. Method: In this paper, we propose a multi-timescale gated neural network for encoding the temporal dependencies from videos. The developed model stacks multiple gated layers in a recurrent pyramid, which makes it possible to hierarchically model not just pairs but long-term dependencies from video frames. Additionally, the model combines the Convolutional Neural Networks into its structure that exploits the pictorial nature of the frames and reduces the number of model parameters. Result: We evaluated the proposed model on the datasets of synthetic bouncing-MNIST, standard actions benchmark of UCF101 and facial expressions benchmark of CK+. The experiment results reveal that on all tasks, the proposed model outperforms the existing approach to build deep stacked gated model and achieves superior performance compared to several recent state-of-the-art techniques. Conclusion: From the experimental results, we can make the conclusion that our proposed model is able to adapt its structure based on different time scales and can be applied in motion estimation, action recognition and tracking, etc.