Research On Improved Seam Carving Image Scaling Algorithm Combining With Saliency Map

The traditional image scaling method does not take into account the important content of the image,but simply changes the size by mechanically scaling the content in the image,which resulting in the lack of edge information and important content of the image.The Seam Carving algorithm is a scaling method that takes into account the important content of the image,which can better preserve the important content and edge information of the image compared with the traditional image scaling method.However,this method simply uses the gradient map to define the energy function.It is easy to cause the obvious structure distortion and the important content missing in the image when there is no clear contrast between the main area of the image and the background color or the main body of the visual part is large.This thesis mainly improves the Seam Carving algorithm from two aspects by introducing the saliency map to change the energy function.First,improvement of Seam Carving algorithm for maintaining the shape and structure of significant objects.Aimed at the phenomenon that the Seam Carving algorithm may destroy the shape and structure of the obvious objects in the image.In this thesis,we use the image weighting method to redefine the significant energy by combining the saliency map and the gradient map of the image.This method not only highlights the important content but also preserves the shape and structure of the obvious objects in the image.The experiment shows that the improved algorithm can better preserve the shape and structure of the obvious objects in the image than the original Seam Carving algorithm.Second,improvement of Seam Carving Image Scaling Algorithm with Visual Significant Graph.Aimed at the phenomenon that it is easy to cause the obvious structure distortion and the important content missing in the image when there is no clear contrast between the main area of the image and the background color or the main body of the visual part is large.This thesis proposes a new visual saliency map combined Seam Carving image scaling algorithm by multiplying the image saliency map and the gradient map as a new energy map of the image.The experiment shows that the improved algorithm can better maintain the important content of the image and achieve a better overall visual effect than the original Seam Carving algorithm.Moreover,the average energy value of the image pixel is increased from 122.74 to 126.54,and which obtains a better scaling quality of the image.This thesis takes Seam Carving algorithm as our research object,mainly improves the lack of the Seam Carving algorithm that can preserve the shape and structure of the obvious objects in the image and the Seam Carving image scaling algorithm combining the visual saliency map.Experiments show that the improved algorithm in this thesis improves the quality of the image when using the Seam Carving algorithm to scaling image.