A5005266091- Image Enhancement Techniques
- Color Science and Applications
- Advanced Vision and Imaging
- Advanced Image Processing Techniques
- Image and Video Quality Assessment
- Image and Signal Denoising Methods
- Advanced Image Fusion Techniques
- Image Retrieval and Classification Techniques
- Color perception and design
- Visual Attention and Saliency Detection
- Visual perception and processing mechanisms
- Advanced Image and Video Retrieval Techniques
- Advanced Chemical Sensor Technologies
- Image Processing Techniques and Applications
- Video Surveillance and Tracking Methods
- Smart Agriculture and AI
- Advanced Neural Network Applications
- Computer Graphics and Visualization Techniques
- Gaze Tracking and Assistive Technology
- Cultural Heritage Materials Analysis
- Metaheuristic Optimization Algorithms Research
- Currency Recognition and Detection
- Spectroscopy and Chemometric Analyses
- Remote-Sensing Image Classification
- melanin and skin pigmentation
University of Milano-Bicocca
2015-2025
University of Milan
2021
This paper reviews the third biennial challenge on spectral reconstruction from RGB images, i.e., recovery of whole-scene hyperspectral (HS) information a 3-channel image. presents "ARAD_1K" data set: new, larger-than-ever natural image set containing 1,000 images. Challenge participants were required to recover hyper-spectral synthetically generated JPEG-compressed images simulating capture by known calibrated camera, operating under partially parameters, in setting which includes...
Assisted living technologies can be of great importance for taking care elderly people and helping them to live independently. In this work, we propose a monitoring system designed as unobtrusive possible, by exploiting computer vision techniques visual sensors such RGB cameras. We perform thorough analysis existing video datasets action recognition, show that no single dataset considered adequate in terms classes or cardinality. subsequently curate taxonomy human actions, derived from...
The authentication process involves all the supply chain stakeholders, and it is also adopted to verify food quality safety. Food tools are an essential part of traceability systems as they provide information on credibility origin, species/variety identity, geographical provenance, production entity. Moreover, these useful evaluate effect transformation processes, conservation strategies reliability packaging distribution flows In this manuscript, we identified innovative characteristics...
This paper presents a review of the NTIRE 2023 challenge on night photography rendering.The goal was to find solutions that process raw camera images taken in nighttime conditions conditions, and thereby produce photo-quality output standard RGB (sRGB) space.Unlike previous year's competition, participants were not provided with large training dataset for target sensor.Instead, this time they given color checker illuminated by known light source.To evaluate results, sufficient number viewers...
This paper reviews the NTIRE 2022 challenge on night photography rendering. The solicited solutions that processed RAW camera images captured in scenes to produce a photo-finished output image encoded standard RGB (sRGB) space. Given subjective nature of this task, proposed were evaluated based mean opinions viewers asked judge visual appearance results. Michael Freeman, world-renowned photographer, further ranked with highest opinion scores. A total 13 teams competed final phase challenge....
Abstract Video restoration concerns the recovery of a clean video sequence starting from its degraded version. Different tasks exist, including denoising, deblurring, super-resolution, and reduction compression artifacts. In this paper, we provide comprehensive review main features existing methods based on deep learning. We focus our attention architectural components, strategies for motion handling, loss functions. analyze standard benchmark datasets use them to summarize performance...
Abstract Annotated datasets for automatic white balance (AWB) are used the evaluation and, when necessary, training, of AWB methods. Relying on such requires awareness potential bias in their content and characteristics: some methods designed to rely presence particular elements, as human skin, while other learn implicit relationships between image light properties from training data. The dependency these makes it fundamental understand whether available actually representative common...
Full-reference image quality measures are a fundamental tool to approximate the human visual system in various applications for digital data management: from retrieval compression detection of unauthorized uses. Inspired by both effectiveness and simplicity hand-crafted Structural Similarity Index Measure (SSIM), this work, we present framework formulation SSIM-like through genetic programming. We explore different terminal sets, defined building blocks structural similarity at levels...
We present a novel Convolutional Neural Network that exploits the Laplacian decomposition technique, which is typically used in traditional image processing, to restore videos compressed with High-Efficiency Video Coding (HEVC) algorithm. The proposed method decomposes frames into multi-scale frequency bands using decomposition, it restores each band ad-hoc designed Multi-frame Residual (MRLN), and finally recomposes restored obtain frames. By leveraging representation of provided by MRLN...
The purpose of this work is to present a new dataset hyperspectral images historical documents consisting 66 family tree samples from the 16th and 17th centuries in two spectral ranges: VNIR (400-1000 nm) SWIR (900-1700 nm).In addition, we performed an evaluation different binarization algorithms, both using single band generating false RGB cube.
We address the task of classifying car images at multiple levels detail, ranging from top-level type, down to specific make, model, and year. analyze existing datasets for classification, identify CompCars as an excellent starting point our task. show that convolutional neural networks achieve accuracy above 90% on finest-level classification This high performance, however, is scarcely representative real-world situations, it evaluated a biased training/test split. In this work, we revisit...
Color constancy algorithms are typically evaluated with a statistical analysis of the recovery angular error and reproduction between estimated ground truth illuminants. Such provides information about only magnitude errors, not their chromatic properties. We propose an Angle-Retaining Chromaticity diagram (ARC) for visual illuminants corresponding errors. provide both quantitative qualitative proof superiority ARC in preserving distances compared to other chromaticity diagrams, making it...
We present a unified deep learning framework for the recognition of user identity and imagined actions, based on electroencephalography (EEG) signals, application as brain-computer interface. Our solution exploits novel shifted subsampling preprocessing step form data augmentation, matrix representation to encode inherent local spatial relationships multi-electrode EEG signals. The resulting image-like is then fed convolutional neural network process dependencies, eventually analyzed through...
We present a state of the art method for vegetable and fruit recognition based on convolutional neural networks. developed our solution around concept smart kitchen/refrigerator equipped with an on-board camera. With this objective in mind, we adopted dataset that was specifically collected annotated according to eating characteristics portrayed items. performed two types experiment: first trained evaluated different state-of-the-art architectures task recognition. Secondly, designed tested...
We focus on saliency estimation in digital images. describe why it is important to adopt a data-driven model for such an illposed problem, allowing universal concept of "saliency" naturally emerge from data that are typically annotated with drastically heterogeneous criteria. Our learning-based method also involves explicit analysis the input at multiple scales, order take into account images different resolutions, depicting subjects sizes. Furthermore, despite training our binary ground...
We propose a novel modular CNN architecture that provides semantic segmentation and understanding of outdoor street environment images. Our solution processes 512x1024 resolution image on single Titan Xp GPU at 37.4 FPS attaining 70.4% IoU the Cityscapes test dataset.
In this paper we present a deep learning method to estimate the illuminant of an image. Our model is not trained with annotations, but objective improving performance on auxiliary task such as object recognition. To best our knowledge, first example architecture for estimation that without ground truth illuminants. We evaluate solution standard datasets color constancy, and compare it state art methods. proposal shown outperform most methods in cross-dataset evaluation setup, competitive...