Search

Blind image denoising and edge-preserving are two primary challenges to recover an image from low-level vision to high-level vision. Blind denoising requires a single denoiser can denoise images with any intensity of noise, and it has practical utility since accurate noise levels cannot be acquired from realistic images. On the other hand, edge preservation can provide more image features for subsequent processing which is also important for the denoising. In this paper, we propose a novel blind universal image denoiser to remove synthesis and realistic noise while preserving the image texture. The denoiser consists of noise network and prior network parallelly, and then a fusion block is used to give the weight between these two networks to balance computation cost and denoising performance. We also use the Non-subsampled Shearlet Transform (NSST) to enlarge the size of receptive field to obtain more detailed information. Extensive denoising experiments on synthetic images and realistic images show the effectiveness of our denoiser. [ABSTRACT FROM AUTHOR]

Image texture description and analysis technology are the basis of many practical applications in pattern recognition. This paper presents a novel and simple, yet powerful method, namely multiple channels local binary pattern (MCLBP), which is the natural extension and development of local binary pattern (LBP) algorithm for color texture representation and classification. MCLBP combines single-channel texture characteristics with multi-channel color information, which reflects the correlations and dependency among different channels. Furthermore, we decompose local color differences into color-difference signs and color-difference magnitudes and MCLBP is extended to MCLBP+M. Then, the resulted image descriptor is a histogram representation, which fuses rich features including color difference sign and color difference magnitude. Comprehensive experiments conducted on five benchmark databases, including Outex, KTH-TIPS, CUReT, STex and KTH-TIPS2-b clearly demonstrate that our proposed method outperforms most of the existed color texture features in terms of classification accuracy. Particularly, our method achieves the best classification performance in CUReT and STex databases. • A multi-channels LBP encodes inner- and inter-channel features of color images. • MCLBP extracts local features from three channels at once. • MCLBP+M characterizes both color difference signs and color difference magnitudes. • Comprehensive experiments are conducted on five benchmark datasets. [ABSTRACT FROM AUTHOR]