Progressive Retinex

Abstract

Contrast enhancement and noise removal are coupled problems for low-light image enhancement. The existing Retinex based methods do not take the coupling relation into consideration, resulting in under or over-smoothing of the enhanced images. To address this issue, this paper presents a novel progressive Retinex framework, in which illumination and noise of low-light image are perceived in a mutually reinforced manner, leading to noise reduction low-light enhancement results. Specifically, two fully pointwise convolutional neural networks are devised to model the statistical regularities of ambient light and image noise respectively, and to leverage them as constraints to facilitate the mutual learning process. The proposed method not only suppresses the interference caused by the ambiguity between tiny textures and image noises, but also greatly improves the computational efficiency. Moreover, to solve the problem of insufficient training data, we propose an image synthesis strategy based on camera imaging model, which generates color images corrupted by illumination-dependent noises. Experimental results on both synthetic and real low-light images demonstrate the superiority of our proposed approaches against the State-Of-The-Art (SOTA) low-light enhancement methods.

Low-Light Image Enhancement by Deep Learning Network for Improved Illumination Map

“…Retinex-Net proposed by Wei et al [15] decomposed and enhanced low-light images through a DecomNet and an EnhanceNet respectively. Wang et al [42] proposed a progressive Retinex model to simulate ambient light and image noise through two separate networks. Moreover, KinD [16]and KinD++ [17] are presented to accomplish image decomposition, denoising, and enhancement via three networks.…”

Assessing the Influence of Sensor-Induced Noise on Machine-Learning-Based Changeover Detection in CNC Machines

“…The spectral slope of Colored noise describes how the power or amplitude of the noise signal varies with frequency and depends on the noise signal and distribution features of noise [ 76 ]. It has a specific non-zero auto-correlation property and is evident in communication systems and image-processing applications [ 77 ]. Periodic Noise : The oscillations and repeating patterns in periodic noise occur over an interval of time.…”

Joint Multi-Scale Tone Mapping and Denoising for HDR Image Enhancement

“…With the advancement in deep learning, learning-based TMOs are become more advantageous, especially in challenging cases. Within the learning-based category, methods can be further divided into supervised [7,49,32,47,6,26,59,42,48,51,12,61,44,38,33,25,22], semi-supervised [52], and unsupervised [57,27,16,60,46,37,23], depending on whether groundtruth is provided during training. In general, most supervised TMOs outperform semi-supervised and unsupervised TMOs, while the latter requires fewer paired training samples and thus can be easily implemented in a wide range of applications.…”