Sanghyun Seo

Directional texture transfer

A texture transfer algorithm modifies the target image replacing the high frequency information with the example source image. Previous texture transfer techniques normally use such factors as color distance and standard deviation for selecting the best texture from the candidate sets. These factors are useful for expressing a texture effect of the example source in the target image, but are less than optimal for considering the object shape of the target image. In this paper, we propose a novel texture transfer algorithm to express the directional effect based on the flow of the target image. For this, we use a directional factor that considers the gradient direction of the target image. We add an additional energy term that respects the image gradient to the previous fast texture transfer algorithm. Additionally, we propose a method for estimating the directional factor weight value from the target image. We have tested our algorithm with various target images. Our algorithm can express a result image with the feature of the example source texture and the flow of the target image.

Extended papers from NPAR 2010: Directional texture transfer with edge enhancement

Texture transfer re-renders a target image with high-frequency information (texture) taken from parts of a reference image that is matched locally to the target image using characteristics such as color. In this paper, we add a directional factor based on the flow of the target image, creating a stroke-like effect that follows edges more accurately. In addition, we propose a method to clearly express object boundaries by considering the effect of the medium on edges in the reference image. We also show how to select an appropriate weight for this directional factor from the reference image. We demonstrate the suitability of this improved form of texture transfer for expressing various artistic styles, and compare our results with those from previous texture-transfer algorithms. We find that our algorithm can be adapted to texture synthesis as well.

A study on stackable mosaic generation for mobile devices

With the growing use of mobile devices such as smartphones, tablet PCs, and digital cameras, photography has become an integral part of everyday life. In particular, mobile devices with cameras allow us to easily capture and modify photographs that can be shared via social networks. In this paper, we introduce a mobile device application for converting photographs into photomosaic images. In general, high-quality photomosaics require large databases. However, mobile devices have limited resources; hence it is difficult for such devices to support photomosaics. We propose a method that produces a photomosaic effect using a database that consists of rotatable images. We also propose a solution to the performance issue based on a best match search.

Interactive painterly rendering with artistic error correction

An artist’s painting is affected by factors such as how he observes objects, his skill in using a brush and materials, and the experience that allows him to correctly apply his skills. The process inevitably results in mistakes contrary to the painter’s original intention. This is a distinguishing factor between painting and photography, but this is the essence of the beauty of painting. The inadequacy of a human being to make a painting exactly as he pleases (as a photograph creates a direct representation of itself) is the starting point of creating a creative work. This paper explains the algorithm that reproduces human errors, as well as the stroke data collection method. Although the results of this research are mainly stylized renderings of modern oil paintings, they have unlimited scalability, in that they can play the role to perform a basic framework. These allow the experimentation with many painting styles through the modification of input data and error generation algorithms.

Color juxtaposition for pointillism based on an artistic color model and a statistical analysis

We analyze characteristics and patterns of color juxtaposition and the color theory used by pointillist painters and employ them to create rendering algorithm that generates images in a pointillist style. We determine the distribution of colors in several paintings by Seurat and quantify a number of the theory that he employs. Using an RYB color wheel and a hierarchical point generation technique, we convert an input image into a set of colored dots, which is again converted into brush strokes with appropriate shapes and directions. We present images which illustrate the extent to which we have managed to simulate Seurat’s technique.

A photomosaic image generation method using photo annotation in a social network environment

With the growing use of social networking services, various applications have been developed to utilize their vast capabilities. Photomosaic techniques, which combine many images to create a new rendering of an input image, can benefit from the capabilities of social networks. In this study, we propose a method that generates a photomosaic image by considering social network context. Our algorithm creates a photomosaic that incorporates photos posted by other users in the users network . We enable the matching function to easily select photos from the albums of users who are connected to the owner of the input image, by computing the closeness of those connections. Moreover, our technique allows the photos in the albums of friends who are annotated in the source image to be matched more effectively.

Image recoloring using linear template mapping

We propose an artistic recoloring method that enhances the perception of complex scenes, while preserving the visual quality of the original image. Our algorithm employs the template which contains the information of color theme. The colors of template is selected based on the Munsell color and are designed on the a ∗ b ∗  chromatic plane of the CIE L ∗ a ∗ b ∗  color space. Template can be used to recolor the original image with simpler and more focused color contrasts. The proposed algorithm consists of the three steps. First, we calculate the regression line for the colors on the original image and the template. Then we create a transformation refer to the relationship between the two regression lines. This transformation is used to change the color distribution in the source image. Finally, the transformed color distribution is mapped from the chromatic a ∗ b ∗  plane to become the template.

A real-time video watermarking algorithm for authentication of small-business wireless surveillance networks

In some business applications, all kinds of cameras sensors are employed in a distributed way to capture videos for different tasks such as surveillance. Once some illegal actions happen, then somebody or some organization wants to forge or replace some surveillance video clips to destroy evidences or obtain illegal profits. How to authenticate the genuineness and integrity of the source video or trace the source of a video information leak becomes a growing requirement in these small businesses. Fortunately, video watermark just provides an effective technology to resolve this issue. This paper proposes a real-time video watermarking scheme for MPEG, where firstly exploits fast scenes segmentation to original video sequence and adaptively selects appropriate scenes to be embedded. Furthermore, visual model is utilized to modulate watermark strength. Watermarks are embedded by adjusting the number of bit1 in the bitstreams through changing level of run-level pairs. Experiment results show little loss of video quality and also exhibit excellent robustness against many attacks. As watermark is directly detected in bitstreams domain, real-time detection becomes a reality. In addition, the embedding strategy guarantees that the bit rate is not increased and the experiments also validate it.

A multi-level depiction method for painterly rendering based on visual perception cue

Increasing the level of detail (LOD) in brushstrokes within areas of interest improved the realism of painterly rendering. Using a modified quad-tree, we segmented an image into areas with similar levels of saliency; each of these segments was then used to control the brush strokes during rendering. We could also simulate real oil painting steps based on saliency information. Our method runs in a reasonable fine and produces results that are visually appealing and competitive with previous techniques.

A painterly rendering based on stroke profile and database

We present a method for producing painting-like images composed of predefined brush strokes. Our proposed method is based on image retrieval method and uses a stroke database. The stroke database consists of transformed copies of the several brush stroke profiles which are obtained from an actual brush stroke. An input image can be reordered in a painterly manner by combination of brush strokes retrieved from the database. Our method is able to produce a painting with diverse media by changing the type of database, such as oil and pastel. In this process, we present a search algorithm to select an appropriate brush stroke from database and an assessment algorithm to judge whether to draw the retrieved brush stroke on the canvas or not. We also introduce a efficient brush stroke model and way of achieving the appearance of thick paint without physical simulation.