Research On Application Of Global Structural Information In Graphics And Image Processing

In computer graphics, the global structural information is the important feature of models and images. It plays an important role in the study and processing. Based on practi-cal problems, we studied some applications of the global structural information in graphics and image processing, including Poisson image editing, color-to-gray conversion, mesh saliency detection and quad remeshing. Due to the consideration of the global structural information, our algorithms get better results than the previous algorithms. This paper con-tributes the following aspects:1. Traditional image compositing methods are used to create images that have plau-sible boundaries. When applied to images of different texture details, these methods will produce unrealistic results. A new framework is proposed. This new framework decompos-es an image to the base layer and the detail layer. The base layer is used to implement the seamless fusion of color at the boundary. The detail layer is used for the smooth transition of texture details at the boundary. We first fuse the base layer, and then synthesize a new detail layer, the new detail layer is added to the fused base layer to get the final result. A weight map is computed by the measure of texture details and distances of pixels to the boundary, which is used to guide the generation of new detail layer. Many experimental results on many images of different texture features show that the proposed algorithm can effectively realize the seamless fusion of color and smooth transition of texture details. The proposed method produces more realistic composites than the traditional method.2. Current color-to-gray methods compute the grayscale result by preserving the dis-criminability among individual pixels. However, human perception tends to firstly group the perceptually similar elements while looking at an image, according to the Gestalt prin-ciples. In this paper, we propose a novel two-scale approach for converting color images to grayscale. First, we decompose the input image into multiple soft segments where each segment represents a perceptual group of content. Second, we determine the grayscale of each perceptual group via a global mapping by solving a quadratic optimization. Last, the local details are added into the final result. Our approach is efficient and provides users quick feedback on adjusting the prominent gray tones of the results. As an important as-pect of algorithm, our approach offers users an easy, intuitive interactive tool for creating art-like black-and-white images from input color images.3. Inspired by basic principles induced by psychophysics studies, we propose a nov-el approach for computing saliency for3D mesh surface considering both local contrast and global rarity. First, a multi-scale local shape descriptor is introduced to capture local geometric features with various regions, which is rotationally invariant. Then, we present an efficient patch-based local contrast method based on the multi-scale local descriptor. The global rarity is defined by its specialty to all other vertices. To be more efficient, we compute it on clusters first and interpolate on vertices later. Finally, our mesh saliency is obtained by the linear combination of the local contrast and the global rarity. Our method is efficient, robust, and yields mesh saliency that agrees with human perception. The algo-rithm tested on many models and outperformed previous works. We also demonstrated the benefits of our algorithm in some geometry processing applications.4. We propose a novel quad remeshing algorithm for articulated shapes. Different from existing methods that are essentially based on local analysis to align with local geometric features, our method focuses on the global structure and generates quad mesh with edge flow aligning with the skeleton. To achieve this, we firstly extract the skeleton from3D shape. It captures the global structure of the shape and decomposes the shape into feature parts. We detect tube-like parts and build sweeping surface on them. For joint part con-necting tube-like parts, we generate a smooth cross field interpolating the directions from its adjacent tube-like parts. Using this cross field, we generate the quad mesh of joint part. It smoothly connects adjacent sweep surfaces. The mesh models generated in our system greatly benefit the sculpting operators for sculpting modeling.