A5090215237- Image Enhancement Techniques
- Color Science and Applications
- Image and Video Quality Assessment
- Advanced Vision and Imaging
- Visual perception and processing mechanisms
- Visual Attention and Saliency Detection
- Advanced Image Processing Techniques
- Advanced Image Fusion Techniques
- Color perception and design
- Image and Signal Denoising Methods
- Image Retrieval and Classification Techniques
- Advanced Image and Video Retrieval Techniques
- Image and Video Stabilization
- Liver Disease Diagnosis and Treatment
- Organ Transplantation Techniques and Outcomes
- Gaze Tracking and Assistive Technology
- Liver Disease and Transplantation
- Image Processing and 3D Reconstruction
- Housing Market and Economics
- Ocular and Laser Science Research
- Advanced X-ray and CT Imaging
- Multisensory perception and integration
- Random lasers and scattering media
- Liver physiology and pathology
- Hybrid Renewable Energy Systems
Computer Vision Center
2007-2024
Universitat Autònoma de Barcelona
2009-2024
Umbo Computer Vision (United Kingdom)
2024
Universitat Pompeu Fabra
2014-2021
University of East Anglia
2019-2021
Society for Imaging Science and Technology
2021
York University
2021
Rochester Institute of Technology
2021
University of Cambridge
2021
University of Quintana Roo
2014-2020
Image dehazing deals with the removal of undesired loss visibility in outdoor images due to presence fog. Retinex is a color vision model mimicking ability Human Visual System robustly discount varying illuminations when observing scene under different spectral lighting conditions. has been widely explored computer literature for image enhancement and other related tasks. While these two problems are apparently unrelated, goal this work show that they can be connected by simple linear...
We propose a novel image-dehazing technique based on the minimization of two energy functionals and fusion scheme to combine output both optimizations. The proposed fusion-based variational (FVID) method is spatially varying image enhancement process that first minimizes previously formulation maximizes contrast saturation hazy input. iterates produced by this are kept, second shrinks faster intensity values well-contrasted regions minimized, allowing generate set difference-of-saturation...
Images obtained under adverse weather conditions, such as haze or fog, typically exhibit low contrast and faded colors, which may severely limit the visibility within scene. Unveiling image structure layer recovering vivid colors out of a single remains challenging task, since degradation is depth-dependent conventional methods are unable to overcome this problem. In work, we extend well-known perception-inspired variational framework for dehazing. Two main improvements proposed. First,...
This paper reviews the NTIRE 2023 challenge on image denoising (σ = 50) with a focus proposed solutions and results. The aim is to obtain network design capable produce high-quality results best performance measured by PSNR for denoising. Independent additive white Gaussian noise (AWGN) assumed level 50. had 225 registered participants, 16 teams made valid submissions. They gauge state-of-the-art
Recent advances in camera designs and imaging pipelines allow us to capture high-quality images using smartphones. However, due the small size lens limitations of smartphone cameras, we commonly find artifacts or degradation processed images. The most common unpleasant effects are noise artifacts, diffraction blur, HDR overexposure. Deep learning methods for image restoration can successfully remove these artifacts. approaches not suitable real-time applications on mobile devices their heavy...
Advancements in imaging technology have enabled hardware to support 10 16 bits per channel, facilitating precise manipulation applications like image editing and video processing. While deep neural networks promise recover high bit-depth representations, existing methods often rely on scale-invariant information, limiting performance certain scenarios. In this paper, we introduce a novel approach that integrates super-resolution architecture extract detailed priori information from images....
Color imaging algorithms - such as color correction, spectral estimation and constancy are developed validated with reflectance data. However, the choice of data set used in development tuning not only affects results these but it also changes ranking different approaches. We propose that this fragility is because difficult to measure/sample enough statistically represent large number degrees freedom apparent reflectances. In paper, we space should be sampled but, rather, integrated....
Finding color representations that are stable to illuminant changes is still an open problem in computer vision. Until now, most approaches have been based on physical constraints or statistical assumptions derived from the scene, whereas very little attention has paid effects selected illuminants final image representation. The novelty of this paper propose perceptual computed corrected images. We define category hypothesis, which weights set feasible according their ability map onto...
Gamut mapping transforms the colors of an input image to a target device so as exploit full potential rendering in terms color rendition. In this paper we present spatial gamut algorithms that rely on perceptually-based variational framework. Our adapt well-known energy functional whose minimization leads enhancement and contrast modification. We show how by varying importance term are able perform reduction extension. propose iterative scheme allows our successfully map from original given...
Color camera characterization, mapping outputs from the sensors to an independent color space, such as \(XYZ\), is important step in processing pipeline. Until now, this procedure has been primarily solved by using a \(3 \times 3\) matrix obtained via least-squares optimization. In paper, we propose use spherical sampling method, recently published Finlayson al., perform perceptual characterization. particular, search for that minimizes three different errors, one pixel based and two...
We propose a method for color stabilization of shots the same scene, taken under illumination, where one image is chosen as reference and or several other images are modified so that their colors match those reference. make use two crucial but often overlooked observations: first, core correction chain in digital camera simply multiplication by 3 × matrix; second, to color-match source we do not need compute matrices, it enough operation transforms matrix into other. This well, which call H....
Visual illusions teach us that what we see is not always represented in the physical world. Their special nature make them a fascinating tool to test and validate any new vision model proposed. In general, current models are based on concatenation of linear non-linear operations. The similarity this structure with operations present Convolutional Neural Networks (CNNs) has motivated study if CNNs trained for low-level visual tasks deceived by illusions. particular, show image denoising,...
The estimation of the illuminant a scene from digital image has been goal large amount research in computer vision. Color constancy algorithms have dealt with this problem by defining different heuristics to select unique solution within feasible set. performance these shown that there is still long way go globally solve as preliminary step In general, evaluation done comparing angular error between estimated chromaticity and canonical illuminant, which highly dependent on dataset. Recently,...
One aspect of the EU funded project SAUCE is to explore possibilities and challenges integrating light field capturing processing into media productions. A special camera was build by Saarland University [Herfet et al. 2018] first tested under production conditions in test "Unfolding" as part project. Filmakademie Baden-Württemberg developed contentual frame, executed post-production prepared a complete previsualization. Calibration post-processing algorithms are Trinity College Dublin Brno...
Abstract The responses of visual neurons, as well perception phenomena in general, are highly nonlinear functions the input, while most vision models grounded on notion a linear receptive field (RF). RF has number inherent problems: it changes with presupposes set basis for system, and conflicts recent studies dendritic computations. Here we propose to model manner, introducing intrinsically (INRF). Apart from being more physiologically plausible embodying efficient representation principle,...
Abstract When light is reflected off a surface, there linear relation between the three human photoreceptor responses to incoming and light. Different colored surfaces have different relations. Recently, Philipona O'Regan (2006) showed that when this singular in mathematical sense, then surface perceived as having highly nameable color. Furthermore, white by corresponding precisely one of four psychophysically measured unique hues. However, O'Regan's approach seems unrelated classical...
Emerging display technologies are able to produce images with a much wider color gamut than those of conventional distribution gamuts for cinema and TV, creating an opportunity the development extension algorithms (GEAs) that exploit full potential these new systems. In this paper, we present novel GEA, implemented as PDE-based optimization procedure related visual perception models, performs (GE) by taking into account analysis distortions in hue, chroma, saturation. User studies performed...
Gamut mapping is the problem of transforming colors image or video content so as to fully exploit color palette display device where will be shown, while preserving artistic intent original content's creator. In particular, in cinema industry, rapid advancement technologies has created a pressing need develop automatic and fast gamut algorithms. this article, we propose novel framework that based on vision science models, performs both reduction extension, low computational complexity,...
We present a color matching method that deals with different non-linear encodings. In particular, given two views of the same scene taken by cameras unknown settings and internal parameters, encoded curves, our is able to correct colors one images making it look as if was captured under other camera's settings. Our based on treating in-camera processing pipeline concatenation matrix multiplication linear image followed non-linearity. This allows us model stabilization transformation among...
This paper presents a virtual laboratory for the computer processing of digital imaging. software application is designed to enhance/reconstruct high-resolution images in order analyze implemented algorithms as well their optical features. The conceived mechatronic, biomedical and electronic engineering students. presented case studies demonstrate accuracy chain used this laboratory, how students could better understand topics related remote sensing, vision, engineering, among others.
Images captured under hazy conditions (e.g. fog, air pollution) usually present faded colors and loss of contrast. To improve their visibility, a process called image dehazing can be applied. Some the most successful algorithms are based on processing methods but do not follow any physical formation model, which limits performance. In this paper, we propose post-processing technique to alleviate handicap by enforcing original method consistent with popular model for haze. Our results upon...