Image Resizing Algorithms Combining Content-aware With Aesthetic Assessment

The traditional image resizing algorithms only consider the size of the target image and ignore the content of the original image in the process of image resizing,easily to cause the loss of important information.Image resizing has been a focus of research in the digital image processing field.This method can identify and protect the important information in be the process of image resizing which can reduce the distortion of image.On the other hand,high aesthetic has been one of the pursuit of goals in the field of computer vision.The computational image aesthetics research attracted more and more attention.Because of the lack of uniform aesthetic rules,the use of computer simulation of human aesthetics evaluate has been one of the challenging problems in the field of computer vision.Existing content-aware resizing algorithms often ignore the impact of the image aesthetics.By analyzing the ideas and features of classic image resizing algorithms,this paper presents and realizes image resizing algorithms combining content-aware with aesthetic assessment.The main works in this dissertation are as follows:(1)By summarizing and generalizing the features and principles of several current image resize algorithms,this paper describes the idea of the algorithm in detail.And we describe several aesthetic features.(2)This paper presents and realizes an aesthetic image resizing algorithm for images containing single object.The algorithm uses image segmentation to improve the weight of important object in the importance map to overcome the deficiencies of traditional methods in the protection of important objects.Meanwhile,we use the aesthetic principles of composition to evaluate the value of image aesthetic.In the process of image resizing,the algorithm improves seam carving and makes the operations of seam caving using the Rule of Thirds.While calculating the importance and the location of the seams,the algorithm improves the aesthetic value of the image by changing the relative distance of the foreground object.The results show that the proposed algorithm,while preserving image information,achieves better performance in terms of aesthetics than a number of other image resizing algorithms.(3)We present and realize an aesthetic image resizing algorithm for landscape images.The algorithm can better detect the importance of image content by combining the visual saliency detection and object detection methods.The visual weight ratio is proposed to evaluate the value of image aesthetic.In the process of image resizing,the algorithm improves seam carving and makes the operations of seam caving using the Rule of golden ratio.The information loss constraint is designed to optimize the selection of seams.The algorithm satisfies the Rule of golden ratio by changing the visual weight ratio of the original image.The results show that the proposed algorithm can achieves better visual balance than a number of other image resizing algorithms.(4)We program the algorithms described above using Matlab,and do experiments with several image resizing algorithms.