《计算机专业英语 教学课件 ppt 作者 霍宏涛 Chapter12》由会员分享，可在线阅读，更多相关《计算机专业英语 教学课件 ppt 作者 霍宏涛 Chapter12（20页珍藏版）》请在金锄头文库上搜索。
1、Professional English in Computer Field,Chapter Twelve Digital Image Processing,内容,正文 A Basic Introduction to Image Processing Adult Image Detection Using Statistical Model and Neural Network 阅读材料 An Introduction to Machine Vision An Overview of JPEG-2000,1 A Basic Introduction to Image Processing,1.
2、 Color Representation 2. Image Representation 3. Histograms 4. Uses of The Histogram 5. The Terminology of Digital Image Processing,Key Words,pictorial adj. 形象化的, 生动的 lay person 非专业人士，门外汉 rudimentary adj. 基本的,初步的 hue n. 色调 saturation n. 饱和度 intensity n. 亮度 cyan n. 青绿色 magenta n. 品红色 pixel n. 像素 blow
3、 up 放大 megapixel n. 百万像素,Key Words,leftmost adj. 最左边 rightmost adj. 最右边 CCD charged coupled device的缩写，电荷耦合器件 histogram n. 直方图 abscissa n. 横坐标 continuous image 连续图像 closed curve 闭合曲线 cumulative distribution function 累积分布函数 probability density function 概率密度函数 quantizing interval 量化间隔 spike n. 毛刺 elusi
4、ve adj. 难以精确定义的,Notes,The inherent subjective appeal of pictorial displays attracts perhaps a disproportionate amount of attention from scientist and lay person alike. It is a broad umbrella under which fall diverse aspects of optics, electronics, mathematics, photography, and computer technology. 人
5、与生俱来的主观上对生动展示的喜爱吸引了从科学家到非专业人士不同的关注，这个广泛的领域包括了光学、电子学、数学、摄影和计算机技术。,Notes,Sampling occurs when you scan an image, or take a picture with a digital camera. As with everything digital, the information provided has to be converted into discrete information, so the scanner/digital camera samples as much in
6、formation as it can and converts it into a digital signal. 当扫描图像或者用数字相机照相时，采样就发生了。和所有的数字化一样，提供的信息必须转化成离散信息，这样扫描仪和数字相机就可以尽量多地获得信息并且转化成数字信号。 The gray-level histogram is a function showing, for each gray level, the number of pixels in the image that have that gray level. 灰度直方图是一个函数，表明在图像中每个灰度级所拥有的像素的数量
7、。,Notes,Notice that under this relatively broad definition of image fall many “representations” that are not perceivable by the eye. 请注意，按照图像的相对广义的定义，许多“表示”是眼睛不可感知的。,2 Adult Image Detection Using Statistical Model and Neural Network,1. Introduction 2. Skin detection 3. Adult image detection 3.1 Feat
8、ure extraction 3.2 Pattern recognition 4. Experiments 5. Summary and conclusions,Key Words,Pornographic adj. 色情图像 Paedophilic adj. 恋童癖 Correlation n. 关联 Probability n. 概率 Semantically 语义的 benign adj. 良性的 entropy n. 熵 feature extraction 特征提取 pattern recognition 模式识别,Key Words,speech recognition 语音识别
9、pixel lattice n. 像素栅格 Methodology n. 方法论 Proposition n. 提议，命题，设想 binarize v. 二值化 morphological adj. 形态的 centroid n. 中心 feature vector 特征向量 feedforward n. 前馈,Key Words,back propagation 反向传播 iteration n. 循环 fluctuation n. 波动 Perceptron n. 感知器,Notes,Detection based on image content analysis has the adv
10、antage to process equally all the images without the need for frequent updating, so will produce more effective filtering. 基于内容分析的图像检测具有如下优势：即同等的处理所有图像，而不需要经常对列表进行更新，因此过滤的效率会更高。 They compared the generalized linear model, the k-nearest neighbor classifier, the multi-layer perception (MLP) classifier
11、 and the support vector machine and found that the MLP gives the best classification performance. 他们比较了归一化线性模型，k邻近法分类，多层感知分类（MLP）和支持向量机，发现MLP具有最好的分类性能。,Notes,In practice we know that the choice of color space is not critical given a histogram-based representation of the color distribution and enough
12、 training data. 在实践中，如果考虑到色彩分布的直方图表示和足够的训练数据，那么色彩空间的选择就不是一个关键问题。 We observed from experiments that for approaches based on skin detection, portraits have a tendence to be detected as adult images since generally portraits expose plenty of skin as adult ones. 通过基于皮肤检测的实验，我们观察到，肖像容易被视作成人图像，因为一般情况下肖像暴露
13、出大量的皮肤区域，这一点与成人图像类似。,Notes,In a successful learning exercise, the system error will decrease with the number of iterations, and the procedure will converge to a stable set of weights, which will exhibit only small fluctuations in value as further learning is attempted. 在一次成功的学习训练中，系统误差会随着循环次数不断降低，并且
14、这个训练过程将会得到一个稳定的权值集合，这个集合在未来的学习训练过程中值只会显示出小的波动。,Reading Material 1 An Introduction to Machine Vision,1. Data size 2. Edge Detection 3. Prototyping 4. Prototyping Algorithm 5. Conclusion,Reading Material 2 An Overview of JPEG-2000,1. Introduction 2. The JPEG-2000 Development Process 3. JPEG-2000 Codin
15、g Engine 3.1 Tiles and Component Transforms 3.2 Partitions, Transforms, and Quantization 3.3 Block Coding 3.4 Packets and Layers,Exercises,The two most commonly used representations are 8-bit greyscale and . 8-bit greyscale contains 256 shades of grey (28 = 256) with 0 normally denoting and 255 deno
16、ting , with other values representing of grey. Simply put, images are stored as collections of , each of which is assigned a color. Of course, in image processing and machine vision, while increased allows for greater detail and more information to work with, it comes at the price of and . If the image is considered a random variable of two dimensions, the area function is proportional to its
