Image Smoothing Base On Spare High Frequency Gradient And Joint Bilateral Filtering

Image smoothing is always a very significant image processing topic.Different kinds of image smoothing algorithms are expected to achieve the same goals:on the one hand,preserving the major structure and edges of the image;on the other hand,suppressing insignificant details.The existing image smoothing algorithms can be divided into two categories according to the different design models:the average-based image smoothing algorithm and the image smoothing algorithm based on the global optimization.These image smoothing algorithms can obtain fine image smoothing performances when smoothing natural images.However,there is a special class of image smoothing problem in the image smoothing research—texture removal.Most of the existing image smoothing algorithms cannot smooth texture images accurately and remove texture patterns in texture images.Few image smoothing algorithms can obtain elementary performances of texture removal,but the main structure and edges cannot be maintained very well in the output images and a few of texture patterns still exist in the output images.This is mainly because the texture contains obvious edges and large gradient changes in the texture images and it is easy to be preserved as the main edges.Once the smoothness of the image smoothing algorithms is increased,the main structure and edges in the image will be smoothed at the same time.Based on the analysis above,this paper proposes a novel image smoothing algorithm based on high frequency gradient and joint bilateral filtering to deal with the texture removal in texture images.First,the high frequency information is used to constrain the input image to get the guided image.The guided image contains less details with the constrain of high frequency information.Second,by combining the guided image with the input image,the improved iterative joint bilateral filtering can obtain the final texture removal result.The improved iterative filter makes the output image achieving better performance in the edge preserving and the texture removal.In order to further improve the performance of the image smoothing algorithm proposed before,this paper continues to improve the image smoothing algorithm,and presents a novel image decomposition algorithm to deal with the input image in advance.The novel image decomposition algorithm can reduce the high frequency information in the input image effectively and make the gradient of the image decomposition result no longer violent.Then the sparse high frequency gradient constraint is used to smooth the intermediate result and the texture removal result is obtained.The image decomposition algorithm greatly improves the effect of the constraint of sparse high frequency gradient,therefore the output image has a better texture removal performance.In the comparison and analysis of the experimental results,first,the method of image smoothing quality evaluation is proposed to evaluate the performance of the image smoothing algorithm in the natural images smoothing and texture removal in texture images.Second,the performances of the two image smoothing algorithms proposed in this paper are compared.Both of the two algorithms can achieve the goal of texture removal,and the improved image smoothing algorithm has better texture removal performance.Third,the improved image smoothing algorithm proposed in this paper is compared with several existing image smoothing algorithms,and the experimental results show that the improved algorithm has a good natural image smoothing effect and a better texture removal performance.Last,some applications of image smoothing are introduced in this paper,which show that the research in this paper has the practical significance.