Search

This paper introduces a robust method for removing objectionable reflection interference in photographs captured through a piece of transparent medium. We exploit the fact that a group of image patches extracted from multiple correlated images with similar transmission lie in a very low-dimensional subspace, leading to a low-rank matrix after patch assembly. This allows us to formulate reflection separation as a per-patch sparse and low-rank decomposition problem which can be well solved by the ALM-ADM strategy. To eliminate the influence of unwanted reflection in patch searching and ensure that the extracted patches has a high similarity regarding their transmission layers, we introduce a new patch similarity metric based on both image intensities and gradients. This improves the performance of reflection separation. In addition, since our method does not require image reconstruction from gradient, color-shifting artifacts can be significantly ameliorated and more scene details can be preserved. Experimental results on both synthetic images and various real-world examples demonstrate that the proposed method achieves high quality reflection separation and performs favorably against many existing techniques.

Most traditional Chinese paintings are painted on hand-made paper that easily suffers from severe spectral changes caused by prolonged light exposure, resulting in color distortion and low contrast. To recover the original appearance of a Chinese painting, especially its background color, an automatic background adjustment framework is proposed. This framework is based on the insightful observation that the fading model of a painting image is analogue to the common hazy image formation model when the painting image is transformed into the K-M (Kubelka-Munk) space. We demonstrate that this fading model is quite useful in extracting pigment lines from any painting image. These pigment lines represent clusters of distinct color pigments used in a Chinese painting, which is the key to density map estimation and background restoration. Experimental results prove that our approach is able to restore a variety of deteriorated Chinese paintings without any user intervention or training.

This paper proposes a real-time system to render with layered materials by using a linearly filterable reflectance model. This model effectively captures both surface and subsurface reflections, and supports smooth transitions over different resolutions. In a preprocessing stage, we build mip-map structures for both surface and subsurface meso-structures via fitting their bumpiness with mixtures of von Mises Fisher (movMF) distributions. Particularly, a movMF convolution algorithm and a movMF reduction algorithm are provided to well-approximate the visually perceived bumpiness of the subsurface with controllable rendering complexity. Then, both surface and subsurface reflections are implemented on GPUs with real-time performance. Experimental results reveal that our approach enables aliasing-free illumination under environmental lighting at different scales.