Research On Visual Feature And Emotion Of Flash Animation Picture

With the rapid development of multimedia and network technology, multimediainformation has gradually become the main form of transmitting data though Internetinformation on high-speed network. Multimedia information including image, audio andvideo information and so on, image is an important form of expression, containing a wealthof emotional information. In recent years ,Flash animation is the popularity in the networkand teaching which is widely used in the multimedia form, containing a large number ofanimated images, and the form and content of the picture are varied, expressing variousemotion, the research on Visual Feature and emotion of flash animation picture becomesthe research subject of this thesis.In this paper the intercepted dynamic pictures of the Flash animation picture are asresearch sample image database, extracting from each image color and texture on low levelvisual feature parameters, designing 16 kinds of affective adjectives(Warm, cheerful,fierce), describing the image expression of emotion, With 0-4 five grades quantizing imagein a certain emotion on performance0 independent, 4 said the full performance, completing the establishment of Flashanimation image sample database、emotional description and quantification of Flashanimation image、the work of extracting low-level visual features on color and texture ofeach Flash animation image, founding the further research on Image classification andretrieval based on emotional semantic.The main jobs are as follows:(1) The establishment of Flash animation image sample database. Due to the study isthe Flash animation picture, first of all, we need to build a image sample databasecontaining various species, large quantity，covering a typical Flash animation picture .Flash animation is often divided into6 major categories, such as MTV, animation, advertising,courseware, greeting cards and games, the common is the MTV, animation, games, fromthe existing Flash animation library, Flash animation based on their expression form andcontent is divided into 6 categories, then according to every kind of Flash animationinformation volume, we use Flash Decompiler Trillix software to intercept images, likecommon MTV, reflecting the relatively large amount of information, and the picturecontent is rich, so the quantity of images interception are more. For each Flash animationimages interception, the principles are the sampling time interval is as short as possible、The image information content is relatively rich、to avoid a similar picture repetition, and atotal of 6 major categories of 2737 interception Flash animation picture constructing theimage sample database to be studied.(2) Extract from each image’s low-level visual feature on color and texture.Image low level visual features include color, texture, shape, etc, Flash animation picturecolor is clear, often by the color rendering of emotional expression, so the color featureextraction is crucial, at the same time on the image texture features are also studied andextraction. With C + + and MATLAB programming to accomplish color feature extractionin HSV color space, and color histogram feature vector in 256 dimensions is as the colorfeature, at the same time we extract texture feature based on Co-occurrence matrix, energy,entropy, moment of inertia, the mean and standard deviation are as the final 8 parameterrepresenting texture feature, making the Low level visual features form.(3)The description and quantitative of Flash animation image emotion. each imagecontains and express different feelings, some emotion performance significantly, some notobvious, in order to comprehensively describe the image expression of emotion, use 16emotional adjectives, including warm, quiet, cheerful, lively, funny, humor, funny,exaggeration, desolate, dull, dreary, chaos, fantasy, adventure, horror, intense, with 0-4 fivegrades quantizing image in a certain degree of emotional adjectives, 0 independent, 1indicated a slight performance, 2 said that the general performance, 3 exhibitedsignificantly, 4 said the full performance, through self assessment and laboratory personnelto help finish of 2737 image emotion quantification, so that each image is mainlyexpressed emotion accurately,...