Flash动画画面的视觉特征与情感研究

发布时间：2018-10-26 21:42

【摘要】：随着多媒体和网络技术的迅速发展，多媒体信息逐渐成为互联网信息高速网络上所传送数据的主要形式。多媒体信息包括图像、音频和视频信息等，图像是一种重要的表现形式，而且蕴涵着丰富的情感信息。Flash动画是近些年来在网络上流行和教育教学中普遍使用的多媒体形式，它包含着大量的动画画面，画面形式、内容多种多样，表达情感丰富，对Flash动画画面的视觉特征与情感的研究就成为本文的研究课题。本文针对Flash动画中的动态画面进行截取作为研究图像的样本库，提取每幅图像的颜色和纹理低阶视觉特征的相关参数，设计用16种情感形容词（温馨、欢快、激烈等）描述图像表达的情感，用0-4五个等级定量图像在某个情感上的表现，0表示无关，4表示充分表现，完成了Flash动画图像样本库的建立、Flash动画图像的情感描述和定量、提取每幅Flash动画图像的颜色和纹理低阶视觉特征工作，为基于图像情感分类和检索进一步研究打下了基础。主要工作如下： (1) Flash动画图像样本库的建立。由于要研究的是Flash动画画面，所以首先需要建立一个涵盖种类较多、数量较大、有一定代表性的Flash动画画面的图像样本库。Flash动画常分为MTV、动画、广告、课件、贺卡、游戏等6大类，其中常见的是MTV、动画、游戏等，，从已有的Flash动画库中将Flash动画根据其表达的内容和形式分为6大类，然后根据每一类Flash动画信息量的大小用Flash Decompiler Trillix软件截取图像，比如常见的MTV，反映信息量比较大，画面内容比较丰富，所以截取图像的数量就多些。针对每一个Flash动画截取图像的原则是采样的时间间隔尽量短、图像信息量相对丰富、避免类似画面的重复等，共截取6大类共2737幅Flash动画画面组成要研究的图像样本库。 (2)提取每幅图像的颜色和纹理低阶视觉特征。图像的低阶视觉特征包括颜色、纹理、形状等，Flash动画画面色彩明显，往往通过色彩的渲染表达情感，所以颜色特征的提取至关重要，同时对图像的纹理特征也进行了研究和提取。用C++和matlab编程实现HSV颜色空间颜色特征提取，颜色直方图特征向量256维作为颜色特征，同时实现基于共生矩阵纹理特征提取，能量、熵、惯性矩、相关的均值和标准差作为最终8个参数表示纹理特征，制成低阶视觉特征表格。 (3) Flash动画图像的情感描述和定量。每一幅图像都包含和表达不同的情感，有的情感表现的明显，有些不明显，为了较全面地描述一幅图像的情感表现，采用16种情感形容词,包括温馨、恬静、欢快、活泼、搞笑、夸张、幽默、有趣、凄凉、枯燥、沉闷、繁乱、虚幻、惊险、恐怖、激烈，用0-4五个等级定量图像在某个情感形容词的程度，0表示无关，1表示略有表现，2表示一般表现，3表示表现明显，4表示充分表现，通过自我评定和实验室人员帮助完成对2737幅图像情感的量化，使得每幅图像主要表达的情感较准确，制成图像情感分析表。
[Abstract]:With the rapid development of multimedia and network technology, multimedia information has gradually become the main form of data transmitted on the Internet information high-speed network. Multimedia information includes image, audio and video information. Image is an important form of expression and contains rich emotional information. Flash animation is a popular multimedia form in recent years and widely used in education and teaching. It contains a large number of animated pictures, picture forms, various content, rich expression of emotion, the study of visual features and emotions of Flash animation picture has become the research topic of this paper. In this paper, the dynamic images in Flash animation are intercepted as the sample library of the study images, the parameters of the low-order visual features of each image are extracted, and 16 kinds of emotional adjectives (warm, cheerful, cheerful) are designed. ) describe the emotion of image expression, use 0-4 grade quantitative image in a certain emotion performance, 0 denote irrelevant, 4 denote full performance, completed the establishment of Flash animation image sample database. The emotion description and quantification of Flash animation images, and the extraction of the low order visual features of each Flash animation image, lay a foundation for the further research of image emotion classification and retrieval. The main work is as follows: (1) Establishment of Flash animation image sample library. Because we want to study the Flash animation picture, we need to establish an image sample database covering a large number of Flash animation pictures. Flash animation is usually divided into MTV, animation, advertisement, courseware, greeting card, etc. There are 6 kinds of games, such as MTV, animation, game, etc. The Flash animation is divided into 6 categories according to its expression content and form from the existing Flash animation library. Then according to the size of each kind of Flash animation information, we use Flash Decompiler Trillix software to intercept the image, such as the common MTV, reflects the large amount of information, the picture content is relatively rich, so the number of captured images is more. In view of the principle of each Flash animation to capture images, the sampling interval is as short as possible, the amount of image information is relatively abundant, and the repetition of similar images is avoided. A total of 2737 Flash animation images of 6 categories are intercepted to form the image sample library to be studied. (2) extracting the low order visual features of each image. The low-order visual features of the image include color, texture, shape and so on. The color of the Flash animation screen is obvious and often expresses emotion through the color rendering, so the extraction of the color feature is very important. At the same time, the texture features of the image are also studied and extracted. C and matlab are used to realize the color feature extraction of HSV color space. The color histogram feature vector 256-dimension is used as the color feature. At the same time, based on co-occurrence matrix texture feature extraction, energy, entropy, moment of inertia, energy, entropy and inertia moment are realized. The associated mean and standard deviation are used as the final eight parameters to represent the texture feature, and the low order visual feature table is made. (3) emotional description and quantification of Flash animation image. Each image contains and expresses different emotions, some of which are obvious and some are not. In order to describe the emotional performance of an image more comprehensively, 16 emotional adjectives are used, including warmth, tranquility, joy and vivacity. Funny, exaggerated, humorous, funny, desolate, boring, dreary, messy, illusory, thrilling, scary, intense, with 0-4 levels of quantitative images on the degree of an emotional adjective, 0 for nothing, 1 for slight performance, 2 for general performance, 3 for obvious performance, 4 for full performance. Through self-assessment and laboratory personnel to help complete the quantification of 2737 images, the main emotions expressed in each image were more accurate. Make image emotion analysis table.
【学位授予单位】：山东师范大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP391.41

【参考文献】