Stroke Based Edge-Aware Tonal Adjustment

With the improvement of living standards,people pursue higher quality of life,digital camera then become popular in our daily life.As the vision basis for us to perceive the world,image is a powerful tool to obtain,express and transport information.Usually,images taken by digital camera need adjustment,for example enhancement and restoration,to achieve a satisfactory effect.Many image processing software have presented editing tools to do various image processing,such as Adobe Photoshop,which is the most common image editing software.However,these tools usually need user strokes to select interesting regions,which make it tedious and time consuming even for professional.Traditional image editing usually adjust the color on the whole image,if there exists any local adjustment,user should select the regions by hand,which require high interaction and take time.When editing the image in a local place,the existing approaches usually can't behave well near the edges.In recent years,somebody presented edit propagation approaches based on user strokes,which propagate the information in user strokes to the entire image by edge aware optimization method.However,these approaches all need to solve a large linear system,which is a computationally complex task,and needs a large memory.The linear system makes it hard to process high dynamic range images.In this paper,we present a novel method to avoid the linear system problem.When editing the local regions in the image,user only need to manually draw some strokes to select the interesting regions on the image,and specifies a corresponding edit to the strokes,applying our algorithm,then we get the local adjusted image.In the stroke based editing algorithm,we need an approach to propagate the stroke to the rest regions of the image,which means that we need an edit propagation approach.Most of the traditional edit propagation approaches get the result by minimizing an energy equation,here we use an improved recursive filter to propagate the user stroke to the rest regions of the image.Recursive filter has an infinite impulse response,which continues the propagation on the image edges.As a result,the regions near the edges are blurred and can't select the similar regions in a correct way.In the recursive filter,we add an adaptive guidance weight to the filter to make the propagation stops when the weight tends to 0.Then we got the adaptive recursive filter named edge aware recursive filter.We use the constructed filter to filter the user strokes to propagate the strokes to the whole image,thus we get the editing map,which has a same size as the input image and record the corresponding editing value.Combining the editing map,user specify parameters and the input image,we use the mapping function to compute the value each pixel should be edit,then we get the final output image.The innovation of this article is that,we construct the adaptive recursive filter.Propagating the stroke to the rest regions by the filter,we can select the similar regions and compute the corresponding value each pixel should be edit at the same time.Our approach doesn't involve linear system during the whole processing,so we have a faster running speed.We get better detail of the image features due to the adaptive edge aware filter.The experimental shows that in the local adjustment of tonal values out approach is better and faster.