Font Size: a A A

Research On The Techniques Of Emotional Image Color Editing

Posted on:2018-11-21Degree:MasterType:Thesis
Country:ChinaCandidate:Y Q JiangFull Text:PDF
GTID:2348330542977403Subject:Computer technology
Abstract/Summary:PDF Full Text Request
Different colors in images convey different emotions,for example,artists often use different color combinations to convey different emotions in their creation.Existing color editing methods can only change the color of the image to change the appearance,cannot accurately meet a given emotion.At present,there are few algorithms for emotional color transfer between images,however,they are not accurate in emotion calculation,resulting in inaccurate emotion transfer of the method.To solve these problems by proposing an emotional color transfer method and an emotional color enhancement method.First of all,in order to obtain more accurate and more natural emotional color transfer result,this paper proposes a high-quality automatic emotion-based color transfer method.In previous methods,emotional calculation methods only take into account the color information of the image and ignore the texture information of the image,which leads to the inaccuracy of the emotion calculation.Therefore,this paper takes these two factors into account in the emotion calculation.In order to find a color combination that can express the target emotion,we build a color-emotion model database,and we also design a new template selection strategy.At last,we propose a color transfer method based on color adjustment and color fusion to ensure the gradient and natural color.Secondly,in order to solve the phenomenon of color blocking,artificial color and anti-natural phenomena in color transfer,this paper proposes an emotion color enhancement algorithm based on deep learning.Our deep net taking into account the image local information,which reflects the image of the semantic information to prevent the phenomenon of anti-natural.In order to make the enhancement results meet user's emotion,we establish the emotion classification network to constrain the color.Finally,the final deep learning framework is obtained by fusing the emotion classification network and the image local semantic feature network.Finally,through the coloring network,the source image for emotional color enhancement.The method presented in this paper has a good effect on natural images.
Keywords/Search Tags:Image editing, Image emotion, Emotion transfer, Image enhancement
PDF Full Text Request
Related items