Separation of reflection components by sparse non-negative matrix factorization

Abstract

Abstract. This paper presents a novel method for separating reflection components in a single image based on the dichromatic reflection model. Our method is based on a modified version of sparse non-negative matrix factorization (NMF). It simultaneously performs the estimation of body colors and the separation of reflection components through optimization. Our method does not use a spatial prior such as smoothness of colors on the object surface, which is in contrast with recent methods attempting to use such priors to improve separation accuracy. Experimental results show that as compared with these recent methods that use priors, our method is more accurate and robust. For example, it can better deal with difficult cases such as the case where a body color is close to the illumination color.

Fast and High Quality Highlight Removal From a Single Image

“…Fig. 8 displays our results in comparison with Akashi et al [26]'s, Shen et al [28]'s and Yang et al [21]'s algorithms. All the images are captured by ourselves using a Nikon D7000 camera with a 50mm f /1.8D lens.…”

“…For images with slight and small scale specularity, such as the 'Apples' scene in the top row, all the four methods can give good separation results. In the 'Butterfly' scene, the chromaticity of the pink wing region (R=0.7715, G=0.3505, B=0.5330) is of high similarity to the normalized white illumination (R=0.5774, G=0.5774, B=0.5774) and there is some color bias in [26]'s, [28]'s and [21]'s results. In contrast, we can still recover the diffuse component correctly.…”

“…Also using the local properties, Kim et al [24] propose an optimization algorithm utilizing the dark channel prior proposed in [25]. Another optimization based method is proposed by Akashi and Okatani [26], who formulate the separation as a non negative matrix decomposition problem. This method is sensitive to the initial values and one must run this method several times to get the most reasonable result.…”

“…For performance comparison with STAR, we compare with the most cited method [21], the newest one [26], and one also based on material clustering [28]. Since the authors of [26] only provide the images on real data, we do not include their results in synthetic experiments.…”

“…Since the authors of [26] only provide the images on real data, we do not include their results in synthetic experiments. Neither source code nor running results of [24] is available, so the comparison is omitted.…”