Research On Application-oriented Image Presentation Technology

Image presentation refers to the organization, transformation and presentation of the image content. Favorable forms of image presentation help to demonstrate the expression of information and it is beneficial to express the atmosphere and mood. Facing different problems, some image presentation technology, such as collage,mosaic, pictorial and Text-overlaid Image etc, combine rules or constraints of human-computer interaction, image processing, Universal design and aesthetics to obtain a better development and realize wider range of application. Image presentation not only contain a single image but also includes the application of multiple images.This paper mainly studies image presentation based on a single image. Due to a number of factors, it is difficult for amateur photographers to take satisfying photos even in some hot place. Faced with the problems of common applications and based on poor low-quality color images, this paper uses a number of images in social media to improve their color, so as to achieve the purpose of color enhancement and improve their Presentation. In view of this idea, this paper proposes a novel approach to recolor given photos via a crowd sourcing based makeover scheme. This approach includs image recommendations and image recoloring.We develop an approach of favorite exemplars suggestion by jointly leveraging contextual and visual information associated with the social images. Usually, photography and editing rules are often dependent on scene and context. Generally processing without regarding to specific contents and contexts may not ensure desired enhancement and realistic effects. Therefore, images similar in content and context should be good examples for recoloring problem of this research. Meanwhile, they should provide users with diverse but possible styles for reference. To achieve this, the paper using GIST features describe the image visual content, meanwhile, using image affiliated GPS information to roughly measure the position information and achieve similar image search by a range-based KD tree. According to dataset of similar image, we use CIE Lab first three order statistics to represent an image's color style approximately and we aim to calculate the visual quality of the image by using image affiliated social information (based on the degree of interest in Flickr image). Eventually, we facilitate Affinity Propagation to obtain image recommendation. Experimental results on paper show that the method implemented on an image search, color measurement and visual quality metrics are feasible, image recommendation meets the supposed expectations.At the same time, we develop a novel superpixel-based recoloring scheme. Since the exemplar is determined by constitute targets, we aims to achieve the goal whose recoloring is to modify the input image's color distribution Ds to match a target color distribution Dt.Because superpixels shave better consistency in spatial and color.meanwhile,it is more robust than the image segmentation, so we use the superpixel based on representation. Integrating superpixel color statistics (mean and variance), texture features and spatial constraints, we utilize soft assignment method to find a better match of superpixel of the input image and the exemplar. On this basis, the color shift values are calculated. Experimental results show that the scheme can achieve better color enhancement on a given exemplar. Compared with common color transfer method, the results proved that the scheme referred in this paper is with high effectiveness.