Jonathan Kervec

Approximate Cross Channel Color Mapping from Sparse Color Correspondences

We propose a color mapping method that compensates color differences between images having a common semantic content such as multiple views of a scene taken from different viewpoints. A so-called color mapping model is usually estimated from color correspondences selected from those images. In this work, we introduce a color mapping that model color change in two steps: first, nonlinear, channel-wise mapping, second, linear, cross-channel mapping. Additionally, unlike many state of the art methods, we estimate the model from sparse matches and do not require dense geometric correspondences. We show that well known cross-channel color change can be estimated from sparse color correspondence. Quantitative and visual benchmark tests show good performance compared to recent methods in literature.

Evaluating the Color Fidelity of ITMOs and HDR Color Appearance Models

With the increasing availability of high-dynamic-range (HDR) displays comes the need to remaster existing content in a way that takes advantage of the extended range of luminance and contrast that such displays offer. At the same time, it is crucial that the creative intent of the director is preserved through such changes as much as possible. In this article, we compare several approaches for dynamic range extension to assess their ability to correctly reproduce the color appearance of standard dynamic range (SDR) images on HDR displays. A number of state-of-the-art inverse tone mapping operators (ITMOs) combined with a standard chromatic adaptation transform (CAT) as well as some HDR color appearance models have been evaluated through a psychophysical study, making use of an HDR display as well as HDR ground-truth data. We found that global ITMOs lead to the most reliable performance when combined with a standard CAT, while more complex methods were found to be more scene dependent, and often less preferred than the unprocessed SDR image. HDR color appearance models, albeit being the most complete solutions for accurate color reproduction, were found to not be well suited to the problem of dynamic range expansion, suggesting that further research may be necessary to provide accurate color management in the context of inverse tone mapping.

A handy calibrator for color vision of a human observer

The variability among human observers is a challenge to the calibration of modern displays based on Light Emitting Diodes (LED) and lasers. The spectra of the displayed colors are so peaky that slight differences in the cone sensitivities in human color vision are sufficient to make two observers perceiving different colors on the same screen. Recently published results give evidence to the existence of a small number of classes of human observers. In this paper, we present first results on the development of a prototype for a lightweight, inexpensive, temporally stable and easy to calibrate and easy to use instrument that allows classification of an observer with normal color vision in a small number of categories. The instrument employs a set of LEDs having specific wavelengths and is controlled by a computer interface. The use of such an instrument will allow adapting displays and viewing conditions to individual observers in color-critical applications.

Correction of over-exposure using color channel correlations

We present a new method for correcting over-exposed areas in images. The method takes advantage of the strong correlation between the RGB color channels and relies on the observation that in images the RGB components are often not over-exposed at the same position. Our solution operates on line profiles, making it ideally suited for hardware implementations. In addition to its low computational complexity, our method can accurately recover information in areas where one or two channels are over-exposed, while it reconstructs information in areas that are fully clipped. We show that our method outperforms previous algorithms through quantitative analysis and we demonstrate an important application of this type of solution in the context of high dynamic range image reconstruction.

Towards an Automatic Correction of Over-Exposure in Photographs: Application to Tone-Mapping

A common artifact in photographs is over-exposure due to bright scene features exceeding the abilities of the camera, and causing image areas to appear flat and lacking in detail. Although a wider luminance range could be captured with HDR techniques, this is often not possible, especially in moving scenes. To address this issue, we propose a novel solution for recovering lost details in clipped and over-exposed areas by taking advantage of channel cross-correlation in RGB images. To automate our approach we propose two improvements: 1) using the image white point, we adaptively estimate a clipping threshold value per image, and 2) to better understand the forms of over-exposure, for an optimal selection of parameters, we construct an image database focusing on over-exposed areas and automatically classify over-exposure as light sources, specular highlights or diffuse surfaces. We evaluate our solution using objective metrics and a subjective study based on an ITU standard protocol, showing that our correction leads to improved results compared to previous related techniques. We explore several potential applications of our method, including extending to video as well as using it as a preprocessing step prior to reverse tone mapping.

Image Selection in Photo Albums

The selection of the best photos in personal albums is a task that is often faced by photographers. This task can become laborious when the photo collection is large and it contains multiple similar photos. Recent advances on image aesthetics and photo importance evaluation has led to the creation of different metrics for automatically assessing a given image. However, these metrics are intended for the independent assessment of an image, without considering the possible context implicitly present within photo albums. In this work, we perform a user study for assessing how users select photos when provided with a complete photo album---a task that better reflects how users may review their personal photos and collections. Using the data provided by our study, we evaluate how existing state-of-the-art photo assessment methods perform relative to user selection, focusing in particular on deep learning based approaches. Finally, we explore a recent framework for adapting independent image scores to collections and evaluate in which scenarios such an adaptation can prove beneficial.

Context-aware clustering and assessment of photo collections

To ensure that all important moments of an event are represented and that challenging scenes are correctly captured, both amateur and professional photographers often opt for taking large quantities of photographs. As such, they are faced with the tedious task of organizing large collections and selecting the best images among similar variants. Automatic methods assisting with this task are based on independent assessment approaches, evaluating each image apart from other images in the collection. However, the overall quality of photo collections can largely vary due to user skills and other factors. In this work, we explore the possibility of context-aware image quality assessment, where the photo context is defined using a clustering approach, and statistics of both the extracted context and the entire photo collection are used to guide identification of low-quality photos. We demonstrate that our method is able to flexibly adapt to the nature of processed albums and to facilitate the task of image selection in diverse scenarios.