Image Semantic Color Layers Decomposition And Its Application

Exploring and editing colors in images is a common task in visual information processing.graphic design and photographic post-production.In digital image editing,layers organize and composite images and provide an interactive and intuitive access for manipulating individual colors.and thus image editing becomes easier.So,for a given image,it becomes critical to have a meaningful set of layers.However,how to get the high-quality layering remains challenging.The goal of this research is to decompose an image into a meaningful set of layers for editing purpose.We design an illumination insensitive and structure-aware image color layer decomposition method.For this problem.without loss of generality,we follow their basic definitions:a color layer is defined as a homogeneous color segment,while all the decomposed layers from a specific image can mix together by a linear additive mode to reconstruct the given image.There are several common representations of layer organization or overlay,the method of image color layer decomposition in this paper is based on the commonly used additive color mixing model.It mainly solve two problems in image color layer decomposition.On the one hand,to reduce the influence of non-uniform illumination on color appearance,we use the reflectance map obtained by the intrinsic image decomposition to replace the original image,and carry out the subsequent image color layer decomposition on the reflectance component.This strategy reduces the effect of non-uniform illumination on image color layer decomposition to a certain extent.On the other hand,to obtain fine layers,we leverage image specific structure information and enforce it by encoding a structure-aware prior into a novel energy minimization formulation.The designed optimization considers the fidelity,our structure-aware prior and permissible ranges of weight vectors simultaneously.Based on the image color layer decomposition algorithm above,we introduce the image semantic segmentation to image layer decomposition and design an image semantic layer decomposition.Targeted image layering is performed on regions with similar colors and different semantics in the image,which facilitates subsequent editing of local regions,and the image semantic layer decomposition is realized.This work conducted experimental validated from two aspects:qualitative and quantitative evaluation for the above two work.Experiments demonstrate that.compared with the existing advanced methods.our method gets better performance than state-of-art methods in terms of layer color quality,smooth transition.and structure preserving,and we get much refine layers and natural recolored appearance.In addition,the overall editing effect is better when the image is edited for regions with similar colors and different semantics.