CC-mamba: Mamba-based color constancy with illumination prior-guided dynamic feature modulation and wavelet-domain attention mechanism

Color constancy (CC) is an important ability of the human visual system to stably perceive the colors of objects despite considerable changes in the color of the light illuminating them. While increasing evidence from the field of neuroscience supports that multiple levels of the visual system contribute to the realization of CC, how the primary visual cortex (V1) plays role in CC is not fully resolved. In specific, double-opponent (DO) neurons in V1 have been thought to contribute to realizing a degree of CC, but the computational mechanism is not clear. This work builds an electrophysiologically based V1 neural model to learn the color of the light source from a natural image dataset with the ground truth illuminants as the labels. Based on the qualitative and quantitative analysis of the responsive properties of the learned model neurons, this work found that both the spatial structures and color weights of the receptive fields of the learned model neurons are quite similar to those of the simple and DO neurons recorded in V1. Computationally, DO cells perform more robustly than the simple cells in V1 for illuminant prediction. Therefore, this work provides computational evidence supporting that V1 DO neurons serve to realize CC by encoding the illuminant, which provides a compelling alternative to the more common hypothesis that V1 contributes to CC by discounting the illuminant using its DO cells. This evidence is expected to not only help resolve the visual mechanisms of CC, but also provide inspiration to develop more effective computer vision algorithms.

Boosting illuminant estimation in deep color constancy through brightness robustness enhancement

Enhancing multi-illuminant color constancy through multi-scale estimation and high-frequency preservation

CSNet: A content and structure-aware approach for color constancy

Maintaining color constancy in polymer extrusion processes is a key difficulty in manufacturing applications, as fluctuations in processing parameters greatly influence pigment dispersion and the quality of the finished product. Preliminary historical data mining analysis was conducted in 2009. This work concentrates on Opaque PC Grade 5, which constituted 2.43% of the pigment; it contained 10 PPH of resin2 with a Melt Flow Index (MFI) of 6.5 g/10 min and 90 PPH of resin1. It also employs a fixed resin composition with an MFI of 25 g/10 min. This research identified the significant processing parameters (PPs) contributing to the lowest color deviation. Interactions between processing parameters, for the same color formulation, were analyzed using statistical methods under various processing conditions. A principle-driven General Trends (GT) diagnostic procedure was applied, wherein each parameter was individually varied across five levels while holding others constant. Particle size distribution (PSD) and colorimetric data (CIE Lab*) were systematically measured and analyzed. To complete this, correlations for the impact of temperature (Temp) on viscosity, particle characteristics, and color quality were studied by characterizing viscosity, Digital Optical Microscopy (DOM), and particle size distribution at various speeds. The samples were characterized for viscosity at three temperatures (230, 255, 280 °C) and particle size distribution at three speeds: 700, 750, 800 rpm. This study investigates particle processing features, such as screw speed and pigment size distribution. The average pigment diameter and the fraction of small particles were influenced by the speed of 700–775 rpm. At 700 rpm, the mean particle size was 2.4 µm, with 61.3% constituting particle numbers. The mean particle size diminished to 2 µm at 775 rpm; however, the particle count proportion escalated to 66% at 800 rpm. This research ultimately quantifies the relative influence of particle size on the reaction, resulting in a color value of 1.36. The mean particle size and particle counts are positively correlated; thus, reduced pigment size at increased speed influences color response and quality. The weighted contributions of the particles, 51.4% at 700 rpm and 48.6% at 800 rpm, substantiate the hypothesis. Further studies will broaden the GT analysis to encompass multi-parameter interactions through design experiments and will test the diagnostic assessment procedure across various polymer grades and colorants to create robust models of prediction for industrial growth. The global quality of mixing polycarbonate compounding constituents ensured consistent and smooth pigment dispersion, minimizing color streaks and resulting in a significant improvement in color matching for opaque grades.

Computer vision and deep learning are essential in diverse fields such as autonomous driving, medical imaging, face recognition, and object detection. However, enhancing low-light remote sensing images remains challenging for both research and real-world applications. Low illumination degrades image quality due to sensor limitations and environmental factors, weakening visual fidelity and reducing performance in vision tasks. Common issues such as insufficient lighting, backlighting, and limited exposure create low contrast, heavy shadows, and poor visibility, particularly at night. We propose QWR-Dec-Net, a quaternion-based Retinex decomposition network tailored for low-light image enhancement. QWR-Dec-Net consists of two key modules: a decomposition module that separates illumination and reflectance, and a denoising module that fuses a quaternion holistic color representation with wavelet multi-frequency information. This structure jointly improves color constancy and noise suppression. Experiments on low-light remote sensing datasets (LSCIDMR and UCMerced) show that QWR-Dec-Net outperforms current methods in PSNR, SSIM, LPIPS, and classification accuracy. The model’s accurate illumination estimation and stable reflectance make it well-suited for remote sensing tasks such as object detection, video surveillance, precision agriculture, and autonomous navigation.

Abstract The human visual system achieves color constancy, allowing consistent color perception under varying environmental contexts, while also being deceived by color illusions, where contextual information affects our perception. Despite the close relationship between color constancy and color illusions, and their potential benefits to the field, both phenomena are rarely studied together in computer vision. In this study, we present the benefits of considering color illusions in the field of computer vision. Particularly, we introduce a learning-free method, namely multiresolution color constancy , which combines insights from computational neuroscience and computer vision to address both phenomena within a single framework. Our approach performs color constancy in both multi- and single-illuminant scenarios, while it is also deceived by assimilation illusions. Additionally, we extend our method to low-light image enhancement, thus, demonstrate its usability across different computer vision tasks. Through comprehensive experiments on color constancy, we show the effectiveness of our method in multi-illuminant and single-illuminant scenarios. Furthermore, we compare our method with state-of-the-art learning-based models on low-light image enhancement, where it shows competitive performance. This work presents the first method that integrates color constancy, color illusions, and low-light image enhancement in a single and explainable framework.

ABSTRACT Although often overshadowed by his achievements in physics and mathematics, Isaac Newton's insights into color perception laid the groundwork for modern perceptual psychology. He proposed that color is not an inherent property of light but an internally generated percept, anticipating key principles in perceptual psychology. While early philosophers like Aristotle discussed perception philosophically, scientific study did not actually begin until the 19th century with figures like Wilhelm Wundt and Hermann von Helmholtz and major concepts like the doctrine of specific nerve energies or labeled line theory. Critics argue that Newton's ideas simply reinforced dualism. Modern neuroscience, however, has confirmed his view that perception arises from internal neural interpretation of stimuli. Phenomena like simultaneous color contrast and color constancy were discovered as early demonstrations of these maxims. Like Darwin in biology, Newton provided a unifying principle that shaped psychology, emphasizing the brain's role in constructing perceptual reality.

ADFNeT: Adaptive decomposition and fusion for color constancy

ABSTRACT The development of imaging technologies allows personal electronic devices to have telephoto and close‐up cameras. The images captured by these cameras sometimes are dominated by a single color (i.e., pure color images). In our earlier work, the PolyU Pure Color dataset was collected and the Pure Color Constancy (PCC) method was proposed, which was the first study investigating color constancy for pure color images. In order to make the method more robust, especially when the images are not as extreme as those included in the PolyU Pure Color dataset, an edge‐aware Pure Color Constancy (ePCC) method is proposed in this article. It adopts a similar architecture as the PCC method, with four additional color features derived from the edge map of an image as the input. Moreover, the PolyU Pure Color dataset V2 was collected. It includes 1271 images, which are not as extreme as those in the PolyU Pure Color dataset V1 and cover a wider range of illuminant colors. The proposed ePCC method was found to have better performance than the PCC method on pure color images, reducing the angular error by 10% with slight increases in the number of parameters and computational resources. Moreover, the ePCC method also resulted in comparable performance to the various state‐of‐the‐art learning‐based methods on images of normal scenes.

In order to achieve color constancy, the visual system needs to estimate the illuminant by referring to the chromatic distribution information in scenes where direct cues to the illuminant were absent. In this study, color constancy was investigated by an achromatic setting with short illuminant durations in five kinds of scenes with different numbers of colors and colored patches. Three of these scenes contained 8, 24, and 96 patches with different colors, and two other scenes contained 96 patches with 8 and 24 colors, respectively. All five scenes had identical space-averaged means, but different variances. The results showed that the color constancy index decreased as the variance of scene colors increased. This indicates that the effect of the number of colors and patches on color constancy was dependent on the scene variance they produced: a greater number of colors and patches tended to generate higher variance, which in turn led to lower color constancy indices. The results suggest that color constancy under brief exposure to multicolor scenes cannot be fully explained by models based on adaptation to the illuminant or mean chromaticity of the scene. Instead, the distribution of colors around the mean also plays an essential role.

The captured underwater images often have color distortion and blur due to the propagation characteristics of light in water. The enhancement for underwater images using deep learning has shown promising prospect. However, most neural networks model the inputs and outputs end-to-end directly, which fails to fully utilize the inherent information present in underwater images. We focus more on the real color distribution in water that contains comprehensive information, and fully leverage these information to restore the underwater images without the color deviation. In this work, an underwater image enhancement model combining Lab color space and double-opponency mechanism is proposed. Lab color space can extract luminance and chroma, which can express wider chroma range based on human visual perception. Double-opponent mechanism can extract color constancy features based on the biological vision mechanism when the lights have attenuation differences. Moreover, an adaptive light estimation module is designed to learn the light map from the outputs of double-opponency to adjust the overall color of the images. Extensive experiments demonstrate that our approach achieves outstanding results in enhancing color and clarity in underwater images.

Color constancy, the ability to perceive consistent object colors under varying illumination, is a core function of the human visual system and a persistent challenge in machine vision. Retinex theory models this process by decomposing an image S into reflectance (R) and illumination (I) components (S′=RI). However, conventional Retinex methods suffer from key limitations: independent RGB processing that disrupts inter-channel correlations, weak grounding in color perception models, non-invertible decomposition (S′≠S), and limited biological plausibility. We propose QRetinex-Net, a unified Retinex framework formulated in the quaternion domain—S=R⊗I, where ⊗ denotes the Hamilton product. Representing RGB channels as pure quaternions enables holistic color processing, biologically inspired modeling, and invertible image reconstruction. We further introduce the Reflectance Consistency Index (RCI) to quantitatively assess illumination invariance and reflectance stability. Experiments on low-light crack detection, infrared–visible fusion, and face detection under varying lighting demonstrate that QRetinex-Net outperforms RetinexNet, KIND++, U-RetinexNet, and Diff-Retinex, achieving up to 11% performance gains, LPIPS ≈ 0.0001, and RCI ≈ 0.988.

Background/Objectives: Diabetic retinopathy (DR) segmentation faces critical challenges from domain shift and false positives caused by heterogeneous retinal backgrounds. Recent transformer-based studies have shown that existing approaches do not comprehensively integrate the anatomical context, particularly training datasets combining blood vessels with DR lesions. Methods: These limitations were addressed by deploying a DeepLabV3+ framework enhanced with more comprehensive anatomical contexts, rather than more complex architectures. The approach produced the first training dataset that systematically integrates DR lesions with complete retinal anatomical structures (optic disc, fovea, blood vessels, retinal boundaries) as contextual background classes. An innovative illumination-based data augmentation simulated diverse camera characteristics using color constancy principles. Two-stage training (cross-entropy and Tversky loss) managed class imbalance effectively. Results: An extensive evaluation of the IDRiD, DDR, and TJDR datasets demonstrated significant improvements. The model achieved competitive performances (AUC-PR: 0.7715, IoU: 0.6651, F1: 0.7930) compared with state-of-the-art methods, including transformer approaches, while showing promising generalization on some unseen datasets, though performance varied across different domains. False-positive returns were reduced through anatomical context awareness. Conclusions: The framework demonstrates that comprehensive anatomical context integration is more critical than architectural complexity for DR segmentation. By combining systematic anatomical annotation with effective data augmentation, conventional network performances can be improved while maintaining computational efficiency and clinical interpretability, establishing a new paradigm for medical image segmentation.

Abstract This study assessed the relative effects of metameric (preserving illumination chromaticity) vs. chromatic illumination changes on the perceived quality of real fruits. 20 normal trichromats viewed four singly-presented fruits under three neutral metameric illuminations (differing in spectral power distribution with constant CCT of ∼6500K) and two broadband chromatic illuminations (∼2000K and ∼10000K) inside a white-walled lightroom illuminated by spectrally tuneable lamps. Participants’ ratings of nine fruit attributes were used to calculate Positive Attributes Scores (PAS) for each fruit-illumination pair. For two fruits (banana and pear), PAS varied significantly under metameric illumination changes, but not under chromatic illumination changes. PAS varied significantly under both metameric and chromatic illumination changes for orange, and under chromatic illumination changes only for red apple. The findings indicate that metameric illumination changes alter perceived food attributes via changes in fruit colour appearance, while changes in illumination chromaticity may influence perceived food quality independently of fruit colour appearance which is largely maintained via colour constancy.

This paper introduces an advanced Underwater Image Enhancement (UIE) framework that integrates multi-level color correction with multi-scale restoration to address the challenges of color distortion and image quality degradation in underwater environments caused by complex lighting variations and haze effects. Central to the framework is the Multi-Level Color Correction Model with Adaptive Factor (MLCC-AF), which leverages the principle of color constancy to estimate light color across multiple levels. The model dynamically adjusts global color balance, corrects local highlight regions, and redistributes channel color energy through an adaptive correction factor, effectively mitigating color deviations and significantly enhancing color fidelity and visual quality. Complementing this, the Multi-Scale Joint Restoration Network (MJRN) and Residual-Based Detail Enhancement Network (RDEN) are proposed to tackle haze effects and recover lost details. MJRN optimizes the dehazing process through joint parameter estimation, while RDEN adaptively enhances critical image features, ensuring superior clarity and detail preservation. Extensive experiments conducted on both reference and non-reference underwater image datasets demonstrate that the proposed method consistently outperforms existing state-of-the-art approaches in terms of color correction, contrast enhancement, and detail restoration. The results underline the method’s efficiency and robustness, offering a promising solution for UIE applications.

Abstract This paper addresses the challenge of accurately estimating both the full illuminant spectral power distribution (SPD) and the per-pixel spectral reflectance from an RGB image captured with a known camera. Full spectral information allows us to perform more accurate white balance, or render the scene under another illuminant. Traditional color constancy methods focus on predicting the illuminant color as a 3-dimensional projection of the infinite-dimensional illuminant SPD onto the camera spectral sensitivity functions (SSFs) space. Because different illuminants can have the same projection in the 3-dimensional SSFs space, those traditional methods cannot differentiate between such illuminants (metamers in the camera response space) and hence remove the color cast resulting from the illumination. We reconstruct the spectral information using a neural network with two interconnected branches: one branch predicts the illuminant SPD, and the other reconstructs the per-pixel spectral reflectance by integrating the predicted SPD within its intermediate layers. Experimental results demonstrate that our framework achieves superior performance in estimating both the illuminant SPD and the per-pixel spectral reflectance compared to the previous approach.

Variations in illumination conditions critically degrade color fidelity in digital images, thereby compromising the accuracy of downstream computer vision tasks. Building upon these historical principles, this paper proposes a self-attention autoencoding feature support vector regression algorithm. The method extracts probability distributions in the luminance-red-green color space as primitive features, reconstructs them through a self-attention augmented autoencoder, and deploys support vector regression for illumination estimation. Experimental validation demonstrates superior robustness against noise and illumination diversity compared to feature-based alternatives. On the linear GreyBall SFU dataset, our method achieves an average 64.4% reduction across six key error metrics relative to the minimum values from all eight comparative methods. On the Cube++ dataset, it yields an average 44.9% reduction across six error metrics relative to the minimum values from all six comparative methods.

The contributions of surface reflectance and incident illumination are entangled in the light reflected to the eye. Historically, the extent to which the perception of one determines the other has long been debated, particularly in empirical studies of surface lightness and color constancy. Despite enormous progress in physical measurements of the spatial, spectral, and temporal properties of natural illumination, and in the ability to generate and control in real time artificial light of an almost infinite variety of spectra, the questions of whether and how people perceive the illumination as a distinct entity with its own color, and the interdependence of perceived surface color on perceived illumination, remain open. Given the rise in novel lighting interventions that modulate illumination spectra in order to improve health, well-being, productivity, and culture, it has become increasingly important to understand the two-way interaction between the visual and nonvisual sensing of illumination.