visual perception
In the early stages of perception, visual analysis is complicated by different appearances of the same item, due to factors like: illumination, type of camera, presence of other objects, possibility of occlusions, motion, etc.. To tackle these problems, early vision and low-level analysis does not focus on semantic objects but instead on the picture as a whole.
Retinex principles derive from a series of experiments demonstrating that human color vision is a local spatial process. The human color sensation does not depend exclusively on the spectral properties of the observed point but also on those of the surrounding regions. Due to this spatial dependency of the color sensation, the human vision system is able to lower gradients and enhance edges while partially removing possible color dominants due to the ambient light. In this activity we study algorithms that emulate the human vision with applications for pattern recognition, object tracking, image indexing and retrieval. We propose mathematical models for color transfer and illuminant invariant object detection, human inspired computational models for color constancy, color correction, and visual image enhancement.
Furthermore, we predict perceptual subjective properties of images and scenes developing deep models that capture characteristics such as aesthetic qualities or feeling of safety.
CONTRAST
The image contrast is a feature capturing the variation of the image signal across the space. Such a feature is very useful to describe the local image structure at different scales and thus it is relevant to many computer vision applications, like image/texture retrieval and object recognition. We study and develop measures of image contrast derived from the Retinex theory and compare them with other popular contrast measures.
related publication
M. Lecca, A. Rizzi and R.P. Serapioni. An Image Contrast Measure Based on Retinex Principles. IEEE Transactions on Image Processing, 30:3543--3554, 2021
M. Lecca. A Gradient-Based Spatial Color Algorithm for Image Contrast Enhancement. International Conference on Image Analysis and Processing - ICIAP, LNCS 11752, pp. 93-103, 2019
VISUAL ENHANCEMENT
It is well known that one of the most difficult tasks is to find a black cat in a dark room. We develop algorithms to improve the visibility of the details and content of input images by adjusting their colors on the basis of local spatial and visual features processed channel by channel, derived from Retinex-inspired spatial color algorithms. Solutions to this classic problem of low-level computer vision using deep learning approaches are also considered.
related publications
M. Lecca. Backlight and Spotlight Image Enhancement Based on Von Kries Model. SN Computer Science 4(5):680, 2023 [doi]
M. Lecca. A Basic Tool for Improving Bad Illuminated Archaeological Pictures. 18th International Conference on Computer Vision Theory and Applications - VISAPP, February 2023
G. Simone, M. Lecca, G. Gianini and A. Rizzi, Survey of methods and evaluation of Retinex-inspired image enhancers. Journal of Electronic Imaging, 31(6):063055, 2022
M. Lecca, G. Gianini and R.P. Serapioni, Mathematical insights into the original Retinex algorithm for image enhancement. Journal of the Optical Society of America A - Optics Image Science and Vision - JOSA, 39(11):2063-2072, 2022
M. Lecca, F. Poiesi. Performance Comparison of Image Enhancers with and without Deep Learning. Journal of the Optical Society of America A - Optics Image Science and Vision - JOSA, 39(4):610-620, 2022
M. Lecca. Underwater image enhancement by the Retinex inspired Contrast Enhancer STRESS, 17th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - VISAPP, Vol. 4, February 2022
M. Lecca. A Retinex inspired bilateral filter for enhancing images under difficult light conditions. 16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - VISAPP, pp. 76-86, February 2021
M. Lecca, A. Torresani and F. Remondino. Comprehensive evaluation of image enhancement for unsupervised image description and matching. IET Image Processing, 2020
M. Lecca. Generalized Equation for Real-World Image Enhancement by Milano Retinex Family. Journal of the Optical Society of America A - Optics Image Science and Vision, 37(5):849--858, 2020
M. Lecca, A. Torresani, and F. Remondino. On Image Enhancement for Unsupervised Image Description and Matching. International Conference on Image Analysis and Processing - ICIAP, LNCS 11752, pp. 82-92, 2019
M. Lecca. A Gradient-Based Spatial Color Algorithm for Image Contrast Enhancement. International Conference on Image Analysis and Processing - ICIAP, LNCS 11752, pp. 93-103, 2019
M. Lecca, C.M. Modena, and A. Rizzi. Using pixel intensity as a self-regulating threshold for deterministic image sampling in Milano Retinex: the T-ReX algorithm, SPIE Journal of Electronic Imaging, 27(1), 2018
M. Lecca. STAR: A Segmentation-based Approximation of point-based sampling Milano Retinex for Color Image Enhancement. IEEE Transactions on Image Processing, 27(12):5802--5812, 2018
COMPUTATIONAL COLOR CONSTANCY
This research is concerned with estimating a color correction between images observing the same scene but acquired under different illuminants, viewpoints, and/or by different devices. Many computer vision methods rely on color features, which are determined by the intrinsic properties of objects and surfaces, as well as the color of the light sources and spectral sensitivity of the capturing device. Color variations induced by these latter sources should, however, often be discarded in solving the computer vision task at hand. Computational color constancy - i.e. the ability to implement invariance to light source and device - is then a desired feature of computer vision systems for indexing and retrieval, recognition, tracking, and many more.
related publications
M. Lecca and S. Messelodi. SuPeR: Milano Retinex implementation exploiting a regular image grid, Journal of the Optical Society of America A - Optics Image Science and Vision, 36(8):1423-1432, 2019
M. Lecca, A. Rizzi and R.P. Serapioni. GRASS: A Gradient-Based Random Sampling Scheme for Milano Retinex, IEEE Transactions on Image Processing, 26(6):2767-2780, 2017
M. Lecca, A. Rizzi and R.P. Serapioni. GREAT: a gradient-based color-sampling scheme for Retinex, Journal of the Optical Society of America A - Optics Image Science and Vision, 34(4):513-522, 2017
G. Simone, E. Cordone, R.P. Serapioni and M. Lecca. On Edge-Aware Path-based Color Spatial Sampling for Retinex: from Termite Retinex to Light Energy-driven Termite Retinex, SPIE Journal of Electronic Imaging, 26(3), 2017
M. Lecca, A. Rizzi and G. Gianini. Energy-driven path search for Termite Retinex, Journal of the Optical Society of America A - Optics Image Science and Vision, 33(1):31-39, 2016
M. Lecca and A. Rizzi. Tuning the locality of filtering with a spatially weighted implementation of random spray Retinex, Journal of the Optical Society of America A - Optics Image Science and Vision, 10(32):1876-1887, 2015
M. Lecca. A Full Linear 3x3 Color Correction between Images. Journal of Real-Time Image Processing, vol. 10, n. 2, 2015, pp. 219-237
S. Mutlu, S. Rota Bulò and O. Lanz. Exploiting Color Constancy for Robust Tracking Under non-Uniform Illumination. International Conference on Image Analysis and Recognition - ICIAR, 2014
L. Michela. On the von Kries Model: Estimation, Dependence on Light and Device, and Application, Advances in Low-Level Color Image Processing, pp. 95 -135, 2014
M. Lecca and S. Messelodi. Linking the von Kries Model to Wien's Law for the estimation of an Illuminant Invariant Image. Pattern Recognition Letters, Vol. 32, No. 15, pp. 2086-2096, 2011
SCENE PERCEPTION
This activity concerns the development of scene classification algorithms based on Convolutional Neural Networks that mimic human perception and are able to predict perceptual subjective attributes (e.g. memorability, aesthetic qualities or feeling of safety of a scene).
For example, cities' visual appearance plays a central role in shaping human perception and response to the surrounding urban environment. The visual qualities of urban spaces affect the psychological states of their inhabitants and can induce negative social outcomes. Hence, to understand people's perceptions and evaluations of urban spaces we developed deep learning techniques to describe perceptually the image of an urban space.
related publications
A. Siarohin, G. Zen, N. Sebe and E. Ricci. Enhancing Perceptual Attributes with Bayesian Style Generation, Asian Conference on Computer Vision - ACCV, 2018
X. Alameda-Pineda, A. Pilzer, D. Xu, N.Sebe and E. Ricci. Viraliency: pooling local virality, Computer Vision and Pattern Recognition - CVPR, 2017
A. Siarohin, G. Zen, C. Majtanovic, X. Alameda-Pineda, E. Ricci and N. Sebe. How to make an image more memorable? a deep style transfer approach. International Conference on Multimedia Retrieval - ACM ICMR, 2017
L. Porzi, S. Rota Bulò, B. Lepri, E. Ricci. Predicting and Understanding Urban Perception with Convolutional Neural Networks, ACM Multimedia - ACMMM, 2015