• 沒有找到結果。

http://www.cs.sjtu.edu.cn/~shengbin/course/SE/sesite/home.html Important!

N/A
N/A
Protected

Academic year: 2022

Share "http://www.cs.sjtu.edu.cn/~shengbin/course/SE/sesite/home.html Important!"

Copied!
85
0
0

加載中.... (立即查看全文)

全文

(1)

Important!

You get:

 Lecture notes,

 Project information,

 Assignments,

 Grading criteria,

 Contact information,

 And more…

http://www.cs.sjtu.edu.cn/~shengbin/course/SE/sesite/home.html

(2)

摄影的光线分析与拍摄建议

light analysis of photography and shooting

suggestions

(3)

(一)逆光摄影

Backlight photography

(4)

颜色直方图 Color histogram

(5)

调大光圈,容易造成背景虚化

(模糊)调慢快门速度,会降低防抖效 果,产生抖动,拖影

加高感光值。会增大噪点

 Adjustable large aperture, easy to cause the bokeh (fuzzy)

Slow shutter speed, will reduce the anti shake effect, jitter, ghosting

Heightening sensitivity. Will increase the noise

(二)夜间摄影 Night photography

(6)

白斑效果 Effect of leukoplakia

运用闪光灯 Using the flash

曝光不足 VS 过度曝光

Under exposure to VS over exposure

(7)

光线的位置 The light position

顺光 Shun the light

逆光 Backlight

顶光 Top light

(8)

日出前,这样的场景让人感觉很清凉

Before sunrise, this scene let people feel very cool

强烈的阳光使景物的颜色发生了转变

The strong sunlight make scenery color change

光线的颜色 The colour of the light

(9)

光线的角度 The angle of the light

(10)

改变照相机朝向

改变聚焦位置

改变基准光

后期处理

 Change camera orientation Change the focus position Change the reference light Post processing

我们的想法: Our idea:

使用闪光灯;

 改变曝光参数(焦点、测光点、曝光补偿);

 HDR 模式;

使用反光板;

其它技巧。

使用侧逆光

 选用更长的焦段

 选择更远的背景

 减少进入相机的背景光线

反其道而行:剪影效果。

 The use of flash;

Change the exposure parameters (focus, metering point, exposure compensation);

HDR mode;

Using the reflector plate;

Other skills.

Use the backlight

Choose a longer focal length The choice of further background To reduce the background light camera The opposite effect: silhouette.

摄影师的办法: The photographer's way:

初步分析 Preliminary analysis

(11)

关键问题: Key problem

如何结合时间、位置、拍照预览等信息建立全局光照模型 How to combine the time, location, preview and other

information to build the global illumination model

一个可能的方案

: A possible solution

光场( light field )

The light field is a function that describes the amount of light faring in every direction through every point in space. – Wikipedia.com

(12)

大屏手机的另一种操作

Another kind of operation of large screen mobile

phone

(13)

手机屏幕越来越大

More and more mobile phone screen

(14)

A car accident caused by that the driver must use her both hands to unlock the phone since the phone is so large

(15)

Our one hand can cover 4 inches screen but not 4.3inches screen

(16)

大屏手机如何操作? Large screen mobile

phone how to operate?

(17)

我们的想法 Our idea

(18)

技术路线 The technical way

(19)

最接近的现有技术 The closest prior art

http://v.youku.com/v_show/id_XNTI2ODQ2NDg4.html

(20)(21)

照片中的人物 Pose 分割与匹配

The people in the photo Pose segmentation

and matching

(22)

目的与意义 Objective and significance

• 在摄像的时候,往往有可能会遇到不知道怎么摆 Pose 、摆什么

Pose 的情况。本项目的就是为了解决这种尴尬,为摄像者提供相

似 Pose 的匹配信息。

• Many people don’t know how to pose when take a photo.This project is to solve this kind of embarrassing, provide similar information for the photographer, Pose.

目标 Target

• 提取 Pose The extraction of Pose

• 数据库建立 The establishment of the database

• Pose 匹配 Pose matching

(23)

• 提取 Pose The extraction of Pose

• 利用人脸识别获取人脸信息

• Gets the face information using face recognition

• 通过图像分割 Pose

• Image segmentation by Pose

识别与分割

• 匹配 Pose

• Matching pose

• 利用 Pose 形状信息查找相似 Pose

• Using the Pose shape information to find similar Pose

(24)

照片中的背景信息匹配

Background information in the photo

(25)

目的与意义 Objective and significance

• 在摄像的时候,往往会因为有美好的景色而不知如何构图的苦恼。

本项目的就是为了解决这种尴尬,为摄像者提供相似背景信息的优 秀摄影构图建议。

• When take a photo, often because of the beautiful scenery and do not know how the composition of distress. This project is to solve this kind of embarrassing, provides the similar background

information of excellent advice for the photographer photography composition.

目标 Target 提取背景信息 Extraction of background information

• 数据库建立 The establishment of the database

• 背景匹配 Background matching

(26)

背景信息匹配 Background information matching

通过对背景的色彩、纹理分析,查找数据库中拥有相似背景的照片以 提供合理构图建议。

Based on the background color, texture analysis, in order to provide reasonable suggestions have similar composition background photo search database.

(27)

大数据研究项目

Large data research project

裴澍炜

youyoupei2013.com 13915217516

(28)

2 :大数据分析挖掘应用 Large data mining application

1 :大数据查询工具(搜索引擎) Big Data query tools(search engine)

两个可选项目

Two optional item

(29)

大数据查询工具(搜索引擎)

Big Data query tools(search engine)

项目内容: The project content:

研究传统和大数据时代的数据搜索算法,研究现有的开源软件包、算法技 术,提出自己的优化算法、架构,或者针对特定应用场景设计最佳算法,编写 相应的代码应用。

Study on the age of the traditional and large data search algorithm, studies the existing open source software package, algorithm, optimization algorithm is presented, their architecture, or for specific application scenarios to design the optimum algorithm, write code to the corresponding application.

项目开展步骤: The project development steps:

项目前期调研 Project preparation

根据现有算法编码 According to the existing algorithm

code 项目需求、设计

The needs of the project, design

算法优化改进

The improved optimization algorithm

比较分析报告

Comparative analysis report

性能测试

Performance test

(30)

大数据查询工具(搜索引擎)

Large data query tools (search engine)

项目前期调研 Project preparation

数据格式 Data format

结构化 Structured

非结构化 Unstructured

关键字

文本 text

语音 Speech

sound 图像 image

数据源

程序输入

算法

Google等公司公开 发表的论文、专利等

爬虫算法 MR算法

Algorithm

Companies such as Google published papers, patents etc.

Crawler algorithm MR algorithm

(31)

大数据查询工具(搜索引擎)

Big Data query tools(search engine)

可能的创新点: The possible innovation point:

1 、匹配算法 2 、排序算法

3 、与 hadoop 等大数据技术结合的架构算法

4 、针对社交网络、电子商务、物联网等特定需求的应用

1, matching algorithm 2, ranking algorithm

3,Architecture algorithm combined with 3, Hadoop and other large data technology

4, in view of the social network, e-commerce, Internet of

things such as the specific needs of the application

(32)

大数据分析挖掘应用

Large data mining application

项目内容: The project content:

利用互联网上公开的数据集,或者工具采集的数据集,建立数学 模型,分析数据关联性,找出规律和有价值信息,编写物联网、互联 网、智慧城市、智能电网、车联网、电子商务等特定场景下的大数据应 用程序

The use of the Internet on public data sets, or the tool for collecting data set, the establishment of mathematical model, correlation data, find out the rules and valuable information, the preparation of large data applications,

Internet, Internet of things, smart grid, smart city car networking, e-commerce and other specific scenarios.

项目开展步骤: The project development steps:

项目前期调研 Project preparation

确定数据集

To determine the data set

项目需求、设计

The needs of the project, design

设计算法模型

The design algorithm model

比较分析报告

Comparative analysis report

性能测试

Performance test

(33)

大数据分析挖掘应用

Large data mining application

确定数据集: To determine the data set:

国外网站:

Foreign websites: 1 、 Public Data Sets on Amazon Web Services (AWS)

2 、 Yahoo! Webscope

3 、 Konect is a collection of network datasets 4 、 Stanford Large Network

Dataset Collection

(34)

大数据分析挖掘应用

Large data mining application

其它数据集(或者仿真、实际采集等) Other data sets (or simulation, the actual collection)

1 、 Dataset for "Statistics and Social Network of YouTube Videos“

2 、 1998 World Cup Web Site Access Logs 3 、 Page view statistics for Wikimedia projects 4 、 AOL Search Query Logs – RP

5 、 CoPhIR 6 、

(35)

大数据分析挖掘应用

Large data mining application

可能的创新点: The possible innovation point:

1 、数据挖掘算法并行化、分布式处理优化

2 、批计算、流计算等算法模型应用创新

3 、在跨学科领域的新发现、新应用

4 、其它在存储、系统架构、安全、隐私保护、人工智

能、云计算等领域的创新

1, data mining, distributed processing optimization parallel algorithm

2, batch computing, flow computation model of application innovation

3, in the application of new discovery, new interdisciplinary field

4, the other in the storage, system architecture, security,

privacy protection, artificial intelligence, cloud computing

and other areas of innovation

(36)

基于 GPU 加速的视频去模糊

To accelerate video based on fuzzy GPU

谯从彬

(37)

这是个神马项目? What is that?

视频去模糊

简单了说:视频从模糊变清晰

复杂了说:基于清晰块插值的模糊视频帧重建方法

整体来讲:研究领域火,应用价值大

GPU并行加速

图像、视频处理加速处理的主要方式

Video deblurring

Simple said: video from a fuzzy variable clear

Complex said: fuzzy video frame reconstruction method based on interpolation clear block Overall: the research field of fire, application value

GPU parallel acceleration

The main way of image processing, video processing acceleration

(38)

现在? Now?

已有一定视频去模糊程序基础

基本实现

但很慢

利用 GPU 加速有一定基础

早前版本加速效果好

但最近版本没有实现加速

Have a certain video deblurring procedure basis The basic realization

But it is very slow

Using GPU to accelerate a certain foundation

Earlier versions of the acceleration effect is good But the latest version did not achieve acceleration

(39)

懂多少才能做? Can I do it?

图像视频基本编程

Opencv都封装好了,懂 c++ 和百度就可以

GPU并行编程( GLSL )

高级语言?最简单的 C 语言

熟悉已有程序

有难度:我自己写程序别人不好看得懂

Video basic programming

Opencv have a good package, understand c++ and Baidu can GPU parallel programming (GLSL)

High-level language? The most simple C language Familiar with the existing program

Difficulty: I write my own program does not look good to understand others

(40)

有什么难度?

 GLSL 高级功能

◦ 并行更快速有效

 难以并行

◦ 需要对算法做一定改进

 又快又好

◦ 做到实时,你们可以发文章了

(41)

Color to Gray

杜皓

(42)

背景 —— 应用

Background -- Application

(43)

背景 —— 问题

Background -- the problem

Photoshop

CIE Y

(44)

新方法 New method

• 映射函数 The mapping function

(45)

新方法 New method

• 颜色差 Color difference

2 2 2

C L a b

      

| i j | G g g

  

(46)

新方法 New method

• 优化 Optimization

2 ,

(

ij

C )

ij

i j

E G



    

(47)

结果 Result

(48)
(49)

Medical image 医学图像

(50)

基于眼底图像的视网膜 血管分割

Based on the retinal fundus images

Vessel segmentation

(51)
(52)
(53)
(54)

为什么要研究眼底图像 Why study fundus

image ?

(55)

 糖尿病

 糖尿病视网膜病

 中心暗影,视力下降,失明

 早期发现,及时治疗

(56)

出血 Hemorrhage

渗出

Exudation

(57)
(58)
(59)
(60)
(61)

血管提取

Extraction of blood vessel

(62)

血管提取 Extraction of blood vessel

(63)

传统方法 The traditional method

(64)

New

Methods ?

(65)

二维码艺术化 QR code Art

半色调 QR Code 制作

Half tone QR Code production

(66)

QR Code

Quick Response Code

二维空间的矩阵条形码

在 4 个角落的其中 3个有 回 字图案“ ”

具备纠错能力

能被快速识别

Matrix bar code in two dimensional space

In the 4 corners of the 3 of them have a "back" and the word pattern Have the ability of error correction

Can be fast identification

(67)

QR Code

用途 —— 广告宣传 Use -- advertising

(68)

QR Code

缺陷 Defect

黑白两色的组成过于单调

用户对编码内容无法感知

需要其他内容进行说明

The composition of black and white too monotonous Users of the encoded content cannot sense

Other contents that need to be explained

(69)

半色调图像

Halftone image

 通过网点的大小或疏密表达图像层次,色彩的变化

 The expression level of the image through the

network size or density, color change

(70)

目标 Target

 将 QR Code 和半色调图像混合

 完成图片到半色调 QR Code 的转换

QR Code and half tone image blending

To complete the picture conversion to a half tone QR Code

(71)

方法 Method

 生成半色调图像 Generate halftone image

 生成 QR Code Generate QR Code

(72)

方法

 图案替换

(73)

方法

 保证可靠性和规整性

(74)

评估

 界面设计 —— 简单的输入输出

 半色调图像生成

 半色调 QR Code 的合成

 * 生成 QR Code 可读性分析

(75)

谢谢!

Thanks

(76)

头发渲染引擎 Hair rendering engine

(77)

项目目标 The project objectives

完成渲染器架构的搭建

设计 UI ,改善软件的使用体验

实现拖拽,载入模型等功能

实现相关算法 (难度可调)

毛发自动建模、光照渲染、物理模拟

Completed the renderer architecture

The design of UI, improve the use of software experience The realization of functions such as drag and drop, load model The realization of the relevant algorithm (adjustable difficulty) Hair modeling, illumination rendering, physical simulation

(78)

项目计划

The project plan

• 第一阶段:实现基本的软件功能:

– 模型的载入和变更

– 模型的简单光照渲染

– 简单交互:拖拽,缩放等

The first stage: the realization of the basic functions of the software:

Loading and change model

The simple light illumination rendering model

Simple interaction: drag, zoom

(79)

大概是这样的。。。 Like that

(80)

项目计划 The project plan

• 第二阶段:添加基础功能

– 头发的自动生成

– 模型的运动

– 简单的物理模拟

The second stage: add basis functions Automatic generation of hair

The motion model

Simple physical simulation

(81)

大概是这个样子。。 Like that

(82)

项目计划

The project plan

• 第三阶段:高级算法的实现(可选)

– 高质量的头发生成算法

– 高质量的毛发渲染算法

– 精细物理模拟

The third stage: the realization of advanced algorithm (optional) Hair generation algorithm of high quality

The hair rendering algorithm with high quality

Fine physical simulation

(83)

最好变成这样子。。。 The best turn out like this

(84)

此项目较有挑战性,欢迎对图形学有兴趣 的同学积极报名!

This project is challenging, welcome interested in

graphics students actively apply!

(85)

參考文獻

相關文件

"Extensions to the k-Means Algorithm for Clustering Large Data Sets with Categorical Values," Data Mining and Knowledge Discovery, Vol. “Density-Based Clustering in

Then, it is easy to see that there are 9 problems for which the iterative numbers of the algorithm using ψ α,θ,p in the case of θ = 1 and p = 3 are less than the one of the

Although we have obtained the global and superlinear convergence properties of Algorithm 3.1 under mild conditions, this does not mean that Algorithm 3.1 is practi- cally efficient,

In this chapter we develop the Lanczos method, a technique that is applicable to large sparse, symmetric eigenproblems.. The method involves tridiagonalizing the given

Like the proximal point algorithm using D-function [5, 8], we under some mild assumptions es- tablish the global convergence of the algorithm expressed in terms of function values,

• When this happens, the option price corresponding to the maximum or minimum variance will be used during backward induction... Numerical

• Given a (singly) linked list of unknown length, design an algorithm to find the n-th node from the tail of the linked list. Your algorithm is allowed to traverse the linked

• When this happens, the option price corresponding to the maximum or minimum variance will be used during backward induction... Numerical