Content-aware Image Zoom Based On Mesh

Image scaling is a basic part of digital image processing.With the rapid development of network technology,the media information,such as image and video,is increasing exponentially.Therefore,it is more common to modify the size of image or video to apply different display devices.For the traditional image scaling method,ignoring the image content and uniformly scaling the image,the output image has different defects.Therefore,considering the scaling method of image content has gradually become one of the new research hotspots.In order to solve the existed shortcomings in the traditional image scaling technology,the content aware based image scaling technology is proposed.There are three problems to be considered in the research.1)How to protect an image's body area and information.2)How to effectively maintain the geometric structure characteristics of an image.3)How to achieve image scaling in real time.Because the problems above are both key problems and difficult problems in the image scaling field,this paper presents a new solutions after analyzing and studying about them,which makes a better visual results in scaling images.The main work and contributions are as follows:1.Improve traditional saliency detection and straight line detection methods.1)Image saliency detection.The saliency obtained by saliency detection method is optimized,which can validly detect the main region and context of main region.A novel method for determining the elastic coefficient is proposed based on the detected saliency,which can effectively protect the content of image via controlling the grid deformation.2)Features straight line detection.On the basis of traditional straight line detection algorithms,all straight lines in the image are detected.Then,in order to maintain straight line in image zoom,primary feature straight lines are extracted from all detected straight lines by employing a certain strategy.2.A content-aware image scaling method based on spring analogy is proposed.The objective function of controlling mesh distortion is constructed by using spring analogy to achieve image scaling effectively.1)Planar triangular mesh is constructed to overlay theimage,and the image zoom is implemented through adopting the spring analogy method.Elasticity coefficients are employed to control the spring deformation,which not only can keep image salient region effectively but also prevent the distortion of significant objects and feature lines.2)A new function of preserving image characteristic straight lines is presented,which is used to keep the characteristic linear structure and geometric structure features of the image.3)In order to prevent over deflection of triangle side,we employ the direction angle of the edge to constraint mesh deformation.Combining the spring analogy method and the above two constraints,objective function is constructed for controlling spring system deformation.Finally,the scaling image is obtained by applying texture mapping technology.Combining the saliency detection and straight line detection,a novel content-aware image resizing method is proposed to control mesh distortion,which is based on straight line maintenance method and edge preserving constraints.The method can validly preserve the main region of the image and feature straight line.Moreover,the results generated by the proposed method have high visual effect.Comparing with the state-of-the-art visual perception image resizing algorithms,the proposed method can resize all kinds of image efficiently.Moreover,the result has better visual effect.