Research On Visual Feature Of Computer Animation Picture

As one major form of multimedia, computer animation is attracting more and moreattention and interests for its characteristics of rich content and the flexible form in the digitalage. Computer animation is a combination of different elements, and its basic one is the image.A large number of images not only constitute the main body of computer animation, but alsocarry its contents of the computer animation, and reflect its emotion. To study the visualcharacteristics of computer animation pictures, it can simplify the complexity of directresearch, understand its contents and find emotional tendency in computer animation.Based on the basic knowledge of image processing, we studied the visual characteristics ofcomputer animation pictures, and used neural network to classify the emotion of computeranimation pictures. The main work includes:Firstly, we established library of a computer animation pictures. In the thesis, we clarifiedthe structural characteristics of computer animation scenes. Each computer animation containsmultiple logical scenes, while each logical scene also contains a number of visual scenes,which constitutes a two-layer tower structure. We established the picture library of computeranimation, which is consisted of2670pictures (including828game pictures,750MTVpictures,520animation pictures,398courseware pictures, and174advertisement pictures),by extracting the node images of logical scenes and representative frame images of visualscenes.Secondly, we realized the extraction of visual characteristics of computer animationpictures, and extracted their color and texture characteristics. In the color characteristics, weextracted the main colors, its percentage, the average color and local color in HSV color space.In the texture characteristic, we extracted the Robert mean, edge density and edge densityvariance. We established the visual characteristic database corresponding to the name ofpictures’. Thirdly, we determined the emotion model of computer animation pictures. According tothe characteristics of computer animation pictures, it showed16emotions (including8positive emotions: warm, quiet, cheerful, lively, joking, exaggerated, humor, funny;8negative emotions: sad, bald, boring, disordered, fantasy, adventurous, horrific, intense) toform a complete emotion model of computer animation pictures and simplify the model byexperiments. We also got the emotional score by investigation and score, and established adatabase of emotional tendency.Fourthly, we used neural network to classify the emotion in the picture library of computeranimation, and achieved a mapping from visual characteristics to emotional characteristics.We established the relationship between visual characteristics and emotional characteristicsby adjusting the number of the training and test sample library and testing different emotionalfeatures to reduce the training error. Compared with classification results and survey scores,we calculated the accuracy of emotional classification with neural network.Based Windows XP operating system and Microsoft Excel database, we achieved theextraction procedures of visual characteristics in computer animation pictures, learned therelationship between characteristic and emotional data with the neural network toolbox inMATLAB7.0, and adjusted different types of emotion to classify the picture emotion.Research results show that we achieved good expectation for extracting visualcharacteristics of computer animation pictures, and got better accuracy to classify the pictureemotion using neural network.Our research on visual characteristics of computer animation pictures helps analyze andunderstand computer animation, contributes to study its emotion, and promotes widely used incomputer animation.