Search

This paper presents a system to recognize face by a variation of LBPH. We use a method of regression of local binary features to get the landmark of face image whose computational complexity is very low. We utilize these landmark points which can be trained to align the face, to extract the facial features. By calculating the Local Binary Patterns Histogram (LBPH) of these landmark points and its neighborhood pixels, we can extract effective facial feature to realize face recognition. This method can increase the calculating speed of LBPH and also can improve the recognition rate. Finally, we show the experimental results using this method to recognize face.

With the development of urban traffic, the demand of the vehicle detector is increasing gradually. The paper presents a vehicle detecting algorithm based on binocular stereo vision. Firstly, in the delineated ROI(region of interest), foreground objects are detected by target detection and background constructor. Then, the depth information of the foreground area is calculated by the principle of stereo vision and the three-dimensional information of foreground area is mapped to the two-dimensional ground plane. Finally, combining depth and two-dimensional information, we can obtain the parameters of moving vehicle and achieve the function of a vehicle detector. The experiment results tested by different videos show that the algorithm we proposed has high robustness and stability.

This paper introduces a novel method to count people which is applied to the ATM surveillance video. In this system, we chose an image without foreground target as the background image. To reduce the interference, updating the background in real-time is a must, and the foreground image is got by the improved Gaussian background difference between current image and background image, then binarizing and morphology processing. After dilating and eroding, the contours of people are clearly presented. We take the contour as a person if the number of the pixels in the contour is more than a certain threshold, and the number of the contours is the number of people. In the end, we utilize the ratio of width to height of the contours to improve the accuracy.

A new improved Goh’s 3-D wavelet transform(WT) coding scheme is presented in this paper. The new scheme has great advantages including a simple code structure, low computation cost and good performance in PSNR, compression ratios and visual quality of reconstructions, when compared to the other existing 3-D WT coding methods and the 2-D WT-based coding methods. The new 3-D WT coding scheme is suitable for very low bit-rate video coding.

This paper proposes a novel solution to realize super-resolution and de-noising for portrait images. Considering that compressive sensing has a good performance on protecting and extracting information in images, it is involved to improve super-resolution. Image blocking is carried out in the process of establishing over-complete dictionaries. After vectorizing all blocks of training samples with different resolutions, the low-resolution and high-resolution over-complete dictionaries turn out by means of placing vectors by pair and correspondingly. On this basis, sparse coefficients of each low-resolution image can be worked out through measurement and OMP algorithm. Depending upon these coefficients, the desired high-resolution image can be constructed. Additionally a well-chosen sparsity is always an important factor that simplifies calculations and gets rid of noises. The experimental result illustrates the effectiveness and robustness of the novel solution.

The ship detection system is designed to detect ships accurately and stably for the automatic ship management in the wharf. The paper discussed a ship target detection method and system, which includes image preprocessing and delineation of ROI; detection of moving objects based on frame difference; features extraction of targets based on edge, angular point, color information, etc.; and image classification. The experimental results show that this system has good accuracy and stability to meet the real-time processing requirements.

A novel approach for counting people passing through a region where is surveilled by a parallel-fixed binocular camera is proposed in this paper. Firstly, two virtual counting lines are set to obtain four line space-time gray images. Then an integrated algorithm based on binocular stereovision and regional gray projection is proposed to extract and split overlapped moving objects in complex counting environment. At last, the object matching and counting is accomplished in space-time disparity maps. The counting algorithm takes full advantage of the objects' 2D and 3D information, and is implemented on the DSP hardware system. The practical test results show that the proposed system can operate in real-time and achieve a counting accuracy of over 96% in crowded people environment.

The paper studies color personal identification photo compression based on object segmentation. Using an object pattern recognition method, a personal ID photo is divided into three regions - background, head-and-shoulder region and facial region. Different regions have different image qualities. After adopting the Freeman chain code description and its arithmetic coding of background boundary, background adaptive gray padding, bit-plane shift quality graded compression algorithm, and JPEG-2000 based wavelet transform and EBCOT coding scheme, a good result that is better than the standard JPEG-2000 is obtained.

The authors present a mathematical model of a microcirculation image. The difference Gauss filter is used to reduce image noise. To measure the blood flow in the curved vessel, the authors use an improved Yokoi skeletonizing algorithm to obtain the vessel's midaxis. A model for measuring blood-flow speed using linear gray interpolation is also discussed. >