Research On Image Aesthetic Evaluation Based On Image Classification

Computer image esthetics is an important embodiment of computer forward tointelligent. How to carry on the computer image aesthetic more effectively is worthstudying. The related researches on computer image aesthetics, which exist athome and abroad, mainly focus on piles of the aesthetic characteristics, and theimage content affects the aesthetic standard for image, so not all aesthetic featuresare suitable for all kinds of aesthetic image. The factor can be ignored that imagecategory will affect the transcendental aesthetic. Making improvement in the defectsof the existing researches can improve the accuracy of image aesthetics in theory.This article first summarizes several kinds of the typical image aestheticfeatures which are commonly used at the present, including the global features andlocal features both. Then with respect to the aesthetic evaluation method mostscholars apply all the aesthetic characteristics to their scoring, this article suggeststhe importance of image classification and image content in aesthetics and theexperimental verification that aesthetic features reflect different adaptability ondifferent image categories, therefore, determines the impact that image class makeson the image aesthetic.Through the analysis of the existing focus on image aesthetic characteristics,specialized image classification method of image aesthetics can be established,including the classification of the color and main body. Then the probabilitydistribution function of image categories is given as classification results and we candetermine the effective aesthetic features of each image classification. After thatthrough training labelled image data and calculating the value of the aestheticcharacteristics,finally image aesthetic structure that internal classification is aheadof image aesthetics can be established.