基于图像边缘度的零树压缩最优分解层的选择

在研究零树图像编码方案的基础上,分析了被压缩图像的空间冗余度(称为图像边缘度)与零树压缩结果的关系,提出了零树压缩存在小波最优分解层的概念,给出了不同压缩比、不同小波分解层与不同类型图像的零树压缩结果的关系,并对相应的压缩时间进行了分析。综合实验结果表明:在进行高比特率压缩时,提倡采用三层小波分解;而进行低比特率压缩时,应采用四层小波分解。这一结论可作为进一步研究具有自适应能力的改进零树算法的基础。

基于人眼视觉特性的彩色图像质量评价

图像处理系统的性能优劣的评判往往需要一个合理迅速的图像质量评价算法作为支撑.传统的图像质量评价算法由于没有充分考虑人眼的视觉特性,使得质量评价结果与实际图像的人眼感知质量不符.根据人眼对图像边缘信息非常敏感这一人眼视觉特性,提出一种综合图像边缘和背景相似度的算法(EBS)来评价彩色图像质量,即通过比较失真彩色图像与原始参考图像的边缘以及除边缘之外的背景相似程度最终确定失真图像的质量.应用于由779幅包含五种类型失真的图像质量评价库的实验结果表明,该算法的评价结果相比PSNR,MSSIM,IFC以及基于像素域的VIF等算法与图像的主观评价结果(由DMOS值表示——将背景不同的一组观察者对失真图像的评分进行统计平均后所得到的评价结果)更一致,也即该算法的评价结果更接近图像的实际视觉感知质量.

适用于MPEG-4的高效空域自适应图像后处理算法

为有效抑制重建视频图像中的人为编码噪声,提出了一种适用于MPEG-4的具有5种滤波模式的高效空域自适应图像后处理算法.通过对各种人为编码噪声的特性分析,根据待滤波区域像素的一维变化特性和像素的变化范围,将待滤波边界划分为不同的区域,并结合人类视觉系统(HVS)的视觉特性,对不同区域分别采取各种有效的局部特征自适应的滤波策略.算法采用了与其他视频编码标准中环路滤波算法相似的算法结构,便于进行兼容多标准的视频解码器设计.实验结果表明,本算法可有效抑制多种人为编码噪声,对于各种不同特性和不同比特率的测试序列均有良好的可靠性和稳定性,相比于传统算法,本算法可获得更高的峰值信噪比增益.

ConGrap -Contour Detection Based on Gradient Map of Images

In this paper, the authors present ConGrap, a novel contour detector for finding closed contours with semantic connections. Based on gradient-based edge detection, a Gradient Map is generated to store the orientation of every edge pixel. Using the edge image and the generated Gradient Map, ConGrap separates the image into semantic parts and objects. Each edge pixel is mapped to a contour by a three-stage hierarchical analysis of neighbored pixels and ensures the closing of contours. A final post-process of ConGrap extracts the contour borderlines and merges them, if they semantically relate to each other. In contrast to common edge and contour detections, ConGrap not only produces an edge image, but also provides additional information （e.g., the borderline pixel coordinates the bounding box, etc.） for every contour. Additionally, the resulting contour image provides closed contours without discontinuities and merged regions with semantic connections. Consequently, the ConGrap contour image can be seen as an enhanced edge image as well as a kind of segmentation and object recognition.

Study on the Technology of Image Measurement with High Accuracy based on Sobel Operator

According to the basic problem of industrial production, the paper presents a monocular camera and automatic annotation ranging scheme with a ruler. The scheme eliminates a series of complex operation steps frequently used for image location of the calibration, image correction, it has simple implementation and has successfully realized precise ranging by using digital image processing technology. From the beginning of the analysis of image edge detection, edge detection principle and some common edge detection operators, and proposes an improved multi-scale edge detection algorithm, then according to the particularity of the edge proposed new edge thinning algorithm, finally mark the target point through the feature extraction, calculate the final results of fault location.

A New Feature-Based Image Registration Algorithm

IR （Image Registration） is one of the important operation of image processing system which is the process of aligning two or more images into one coordinate system that are taken at different times, from different sensors, or from different viewpoints. It has a lot of applications especially medical imaging and remote sensing. The main purpose of this paper is to provide a comprehensive review of existing literatures available on image registration system and proposed a new feature-based IR technique using edge of images. We used edges as a feature of images for registration. It will be a useful document for researchers who will work on feature-based image registration regardless for specific applications.

The Study of Image Segmentation Based on the Combination of the Wavelet Multi-scale Edge Detection and the Entropy Iterative Threshold Selection

This paper proposes an image segmentation method based on the combination of the wavelet multi-scale edge detection and the entropy iterative threshold selection.Image for segmentation is divided into two parts by high- and low-frequency.In the high-frequency part the wavelet multiscale was used for the edge detection,and the low-frequency part conducted on segmentation using the entropy iterative threshold selection method.Through the consideration of the image edge and region,a CT image of the thorax was chosen to test the proposed method for the segmentation of the lungs.Experimental results show that the method is efficient to segment the interesting region of an image compared with conventional methods.

Tumor segmentation in lung CT images based on support vector machine and improved level set

In lung CT images, the edge of a tumor is frequently fuzzy because of the complex relationship between tumors and tissues, especially in cases that the tumor adheres to the chest and lung in the pathology area. This makes the tumor segmentation more difficult. In order to segment tumors in lung CT images accurately, a method based on support vector machine(SVM) and improved level set model is proposed. Firstly, the image is divided into several block units; then the texture, gray and shape features of each block are extracted to construct eigenvector and then the SVM classifier is trained to detect suspicious lung lesion areas; finally, the suspicious edge is extracted as the initial contour after optimizing lesion areas, and the complete tumor segmentation can be obtained by level set model modified with morphological gradient. Experimental results show that this method can efficiently and fast segment the tumors from complex lung CT images with higher accuracy.

An adaptive tensor voting algorithm combined with texture spectrum

An adaptive tensor voting algorithm combined with texture spectrum is proposed. The image texture spectrum is used to get the adaptive scale parameter of voting field. Then the texture information modifies both the attenuation coefficient and the attenuation field so that we can use this algorithm to create more significant and correct structures in the original image according to the human visual perception. At the same time, the proposed method can improve the edge extraction quality, which includes decreasing the flocculent region efficiently and making image clear. In the experiment for extracting pavement cracks, the original pavement image is processed by the proposed method which is combined with the significant curve feature threshold procedure, and the resulted image displays the faint crack signals submerged in the complicated background efficiently and clearly.