Xiuzi Ye

A hybrid method for robust car plate character recognition

Image-based car plate recognition is an indispensable part of an intelligent traffic system. The quality of the images taken for car plates, especially for Chinese car plates, however, may sometimes be very poor, due to the operating conditions and distortion because of poor photographical environments. Furthermore, there exist some similar characters, such as 8 and B, 7 and T and so on. They are less distinguishable because of noises and/or distortions. To achieve robust and high recognition performance, in this paper, a two-stage hybrid recognition system combining statistical and structural recognition methods is proposed. Car plate images are skew corrected and normalized before recognition. In the first stage, four statistical sub-classifiers recognize the input character independently, and the recognition results are combined using the Bayes method. If the output of the first stage contains characters that belong to prescribed sets of similarity characters, structure recognition method is used to further classify these character images: they are preprocessed once more, structure features are obtained from them and these structure features are fed into a decision tree classifier. Finally, genetic algorithm is employed to achieve optimum system parameters. Experiments show that our recognition system is very efficient and robust. As part of an intelligent traffic system, the system has been in successful commercial use.

Error resilience schemes of H.264/AVC for 3G conversational video services

The emerging of third generation mobile system (3G) makes conversational video transmission in wireless environment possible, and the latest 3GPP/3GPP2 standards suggest 3G terminals support H.264/AVC. Due to high bit error rate in wireless environment, error resilience schemes are necessary for 3G terminals. Moreover, according to 3GPP/3GPP2, 3G terminals support only part of error resilience tools of H.264/AVC. This paper lists various error resilience tools and analyzes their usability in 3G environment, and at the same time, some error resilience schemes are proposed. Performances of these error resilience schemes are tested using offline common test conditions. Experiments shows that, in 3G environment, encoding with simple FMO mode and extra intra block refreshing can achieve the best error correction ability, because this method can make full use of spatial correlation for intra predication and reduce the requirement of bit rate.

H.264/AVC error resilience tools suitable for 3G mobile video services *

The emergence of third generation mobile system (3G) makes video transmission in wireless environment possible, and the latest 3GPP/3GPP2 standards require 3G terminals support H.264/AVC. Due to high packet loss rate in wireless envi- ronment, error resilience for 3G terminals is necessary. Moreover, because of the hardware restrictions, 3G mobile terminals support only part of H.264/AVC error resilience tool. This paper analyzes various error resilience tools and their functions, and presents 2 error resilience strategies for 3G mobile streaming video services and mobile conversational services. Performances of the proposed error resilience strategies were tested using off-line common test conditions. Experiments showed that the proposed error resilience strategies can yield reasonably satisfactory results.

Uniform color transfer

We present in this paper a general algorithm called uniform color transfer for transferring dominant colors in one image to another. Our algorithm frees users from interactions required for colorizing grayscale image or transferring color between images. Different from existing algorithms, we cluster the source color image into regions using the GMM-EM method, and the target image using the K-means algorithm. We then impose chromatic mean values on corresponding source image regions in the target image. Experiments showed that our algorithm can give better results.

Automatic skin decomposition based on single image

This paper proposes an automatic decomposition algorithm for separating skin image into melanin and hemoglobin components. First, skin region is detected from input image. Then, the whole image is divided into several sub regions. After extracting separation vectors from each sub regions, we use proposed method to distinguish the vectors which corresponds to melanin and hemoglobin components, respectively. Based on our validity rules, invalid vectors are discarded and valid vectors are treated as candidate vectors. Finally, we determine the final separation vectors which have minimal deviation from all of the candidate vectors to separate skin image. The separation results demonstrate the validity of our algorithm. Our experiments with lower resolution and poor quality images as input also verify that the proposed algorithm is very robust.

Radius-normal histogram and hybrid strategy for 3D shape retrieval

Recent development in computer hardware and modeling technologies has led to a fast increasing of 3D models. To help user find expected models accurately and efficiently, technologies on content-based 3D model retrievals have become one of the most challenging research topic recently. In this paper, radius-normal histogram(RNH) is proposed to describe shape contents and used for shape retrievals. The RNH shape descriptor first uses a series of concentric spheres to capture the point distribution information of the given model. Then for points in each concentric sphere, a radian normal angle is computed to extract the local geometry features. Finally, the radius-normal histogram is constructed by using the extracted shape signatures. The proposed shape representation remains invariant under rotations. It can be generated from the given 3D model efficiently and easily as well. Performance comparisons for the shape benchmark database have proven that the proposed algorithm can achieve better retrieving performance than other similar histogram-based shape representations. To further improve the retrieving performance, different shape features are combined based on dynamic weight selections and a hierarchical architecture is used to speed up the matching process.

PENCIL: a framework for expressing free-hand sketching in 3D

This paper presents a framework for expressing free-hand sketching in 3D for conceptual design input. In the framework, sketch outlines will be recognized as formal rigid shapes first. Then under a group of gestures and DFAs(deterministic finite automata) control, the framework can express users free sketching intents freely. Based on this framework, we implemented a sketch-based 3D prototype system supporting conceptual designs. User can easily and rapidly create 3D objects such as hexahedron, sphere, cone, extrusion, swept body, revolved body, lofted body and their assemblies by sketching and gestures.

Human Contour Extraction Using Level Set

Human contour information is important for human activity understanding and human motion capture. Traditional human contour extraction methods are based on image morphological operation, but they fail when the contour information is dim and discontinuous. In this paper, a novel method using level set technique for human contour extraction in video frames with still background is proposed. Firstly the human profile is detected via image difference with background. After appropriate thresholding and morphological operation, the possible human profile region is then found out according to the image projection using the detected region boundary as the initial evolving curve, the accurate human contour is extracted using a modified level set method, experiments show that this method is faster and more efficient than traditional level set methodsIn this paper, a novel adaptive soft morphological Laplacian (ASML) filter is proposed, based on a combination of the ideas of the soft morphological filtering and the Laplacian operator. ASML filter is an adaptive nonlinear edge detector. The adaptivity is achieved by employing four directional structuring elements and the dynamic repetition parameter. ASML filter has the capability of selecting the directional structuring element with the maximum response, whose direction varies depending on the change of directional edges. In addition, by replacing the Laplacian operator with soft morphological one, the ASML filter provides an alternative algorithm for edge detection and noise suppression. The experimental results show that the ASML filter outperforms the traditional edge detector in the two aspects

Highlight detection and removal based on chromaticity

The presence of highlight can lead to erroneous results in Computer Vision applications such as edge detection, and motion tracking. Many algorithms have been developed to detect and remove highlight. In this paper, we propose a simple and effective method for detecting and removal of highlight. We first use a window to help to remove the noise and reduce the data amount for analysis. We then apply K-means algorithm in a 5-D vector space to computer diffuse chromaticity. In the case of non-white illuminant, illuminant chromaticity is estimated in the inverse-intensity space, and we use Fuzzy C-mean clustering and linear fitting to get illuminant chromaticity. Finally, we use Specular-to-Diffuse mechanism to separate specular reflection component from image. Experiments show that it is robust and can give good results.

Slant correction of vehicle license plate image

Because of perspective distortion between the camera and the license plate, slanted images commonly appear in License Plate Recognition System (LPR system), and it seriously affects the recognition result. This paper presents two types of image slant, and gives an efficient way to rectify them. For horizontal slant correction, the proposed method is based on connected areas labeling and straight-line fitting. For vertical slant correction, the method is based on rotation experiments with various angles. Practical use in the LPR system shows that this method has a correct rate above 97%.