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.

Low-light image enhancement is a widely studied field of image processing. Over the past decades, numerous successful algorithms have been proposed to address its challenges, such as low contrast, color fading, and noise. One way to enhance dim scenes with vivid colors while suppressing noise is to apply intrinsic image decomposition prior to enhancement. Intrinsic image decomposition separates an image into its fundamental components, such as reflectance and shading. In this study, we extent our recently proposed learning-free intrinsic image decomposition method based on Retinex theory. This approach utilizes scale-space computations and super-pixel segmentation to effectively estimate the reflectance and shading components of a scene. With slight modifications to the original method, we demonstrate its usability in the context of low-light image enhancement. To validate the effectiveness and generalizability of the method, we conduct comprehensive experiments on five benchmarks. Our results show that our method, originally designed for intrinsic image decomposition, can produce effective results for low-light image enhancement and achieve competitive performance compared to learning-based methods specifically developed for this task.

Abstract The human visual system unconsciously determines the color of the objects by “discounting” the effects of the illumination, whereas machine vision systems have difficulty performing this task. Color constancy algorithms assist computer vision pipelines by removing the effects of the illuminant, which in the end enables these pipelines to perform better on high-level vision tasks based on the color features of the scene. Due to its benefits, numerous color constancy algorithms have been developed, and existing techniques have been improved. Combining different strategies and investigating new methods might help us design simple yet effective algorithms. Thereupon, we present a color constancy algorithm based on the outcomes of our previous works. Our algorithm is built upon the biological findings that the human visual system might be discounting the illuminant based on the highest luminance patches and space-average color. We find the illuminant estimate based on the idea that if the world is gray on average, the deviation of the brightest pixels from the achromatic value should be caused by the illuminant. Our approach utilizes multi-scale operations by only considering the salient pixels. It relies on varying surface orientations by adopting a block-based approach. We show that our strategy outperforms learning-free algorithms and provides competitive results compared to the learning-based methods. Moreover, we demonstrate that using parts of our strategy can significantly improve the performance of several learning-free methods. We also briefly present an approach to transform our global color constancy method into a multi-illuminant color constancy approach.

We recently proposed a color constancy method based on the observations that the human visual system might be "discounting the illuminant" by using space-average color and the highest luminance patches. Based on these observations, our algorithm relies on two assumptions: (i) there are several bright pixels in the scene, and (ii) the world is gray, on average. The main idea of the algorithm is to estimate the illuminant by finding the deviation of the brightest pixels from the gray value. During experiments, we observed that some pixels decrease the performance of the method. In this work, the algorithm is modified to eliminate the impact of these pixels. According to the comprehensive experiments, the proposed method surpasses several existing approaches on two color constancy benchmarks. Also, we show that the performance of some existing color constancy algorithms can be increased by using a block-based approach and salient pixels.

The effects of strong color casts in traditional, learning-based and data-driven color constancy algorithms is analyzed. This is the first study investigating the response of color constancy methods to illuminants on the edges and outside the color temperature curve. According to the comprehensive experiments, while traditional studies do not fail to "discount the illuminant" from inputs which have strong color casts, the efficiency of learning-based and data-driven algorithms in obtaining canonical outputs decreases significantly compared to traditional methods. We discuss the reasons behind this performance decay and introduce a traditional color constancy algorithm, which presents competitive results in a challenging dataset.

The goal of intrinsic image decomposition is to recover low level features of images. Most of the studies tend to consider only reflectance and shading, even though it is known that increasing the number of intrinsics is beneficial for many applications. Existing intrinsic image datasets are quite limited. In this study, a dataset is introduced to provide a comprehensive benchmark to the field of intrinsic image decomposition. IID-NORD contains a large number of scenes and for each scene ground truth reflectance, shading, surface normal vectors, light vectors and depth map is provided to allow detailed decomposition. Moreover, diverse illuminants, viewing angles, and dynamic shadows are used to prevent any bias. To the best of available knowledge, IID-NORD is the most comprehensive dataset in the field of intrinsic image decomposition. IID-NORD will be available on the first author’s official webpage.

Collisions cannot be entirely avoided during normal operation of an autonomous mobile robot. Therefore, mobile robots need to detect collisions and react appropriately when they happen. We investigate whether logistic regression on acceleration data can be used for collision detection. We have collected training data from an acceleration sensor during normal driving behavior of a small mobile robot. Collisions were manually marked by a human operator. Accelerations occurring in a direction opposite to the current direction of motion are more likely to be actual collisions. Hence, we combine accelerometer data and motor commands in the logistic regression model. The trained model was able to detect 13 out of 14 collisions on a separate test set with no false positives.

A general video game player is an an agent that can learn to play different video games with no specific domain knowledge. We are working towards developing a GP-based general video game player. Our system currently extracts game state features from screen grabs. This information is then passed on to the game player. Fitness is computed from data obtained directly from the internals of the game simulator. For this paper, we compare three different types of game state features. These features differ in how they describe the position to the nearest object surrounding the player. We have tested our genetic programming game player system on three games: Space Invaders, Frogger and Missile Command. Our results show that a playing strategy for each game can be found efficiently for all three representations.

This study is concerned with the problem of shadow detection and removal from single images of natural scenes. In this work, the authors propose a shadow detection method with a surface descriptor, termed colour‐shade, which allows them to include the physical considerations derived from the image formation model capturing gradual colour surface variations. The authors incorporate a colour‐shade descriptor into the condition random field model to find same illumination pairs and to obtain coherent shadow regions. The authors propose a shadow removal method using an improved local colour constancy computation, which uses anisotropic diffusion to estimate the illuminant locally for each image pixel in shadow. The authors evaluate their method on two shadow detection databases. The experimental results demonstrate that their shadow detection and removal method is state of the art.

Summary form only given. The ability of the human visual system to perceive colors as approximately constant irrespective of the illuminant is called color constancy. The output of a sensor varies with the type of illuminant used and is therefore not color constant. Color constancy algorithms try to compute a color constant descriptor from the measured data (M. Ebner, 2007). We show how color constancy can be implemented within JPEG and JPEG2000 by an adaptation of the well-known Retinex algorithm (E.H. Land and J.J. McCann, 1971), (E.H. Land, 1980). Compared to the normal decoding processes only little additional overhead is introduced into the processing pipeline.

This paper describes the use of a multi-objective evolution strategy in tuning a fuzzy controller which is used in sewage treatment plants. The controller adjusts the oxygenation in the aeration tank to support biological processes, which reduce nutrients. Globally optimal settings are unknown and up to now the settings are made by a specialist equipped with a large amount of background knowledge. We successfully approximated the pareto-optimal set of solutions considering given constraints and three objectives using a wastewater simulation software. Evolution was able to find the settings proposed by a human expert as well as an interesting parameter niche, the so called simultaneous intermittent denitrification, which promises energy savings.

A digital sensor which is used inside a digital camera usually responds to a range of wavelengths. The response of the sensor is proportional to the product of the irradiance falling onto the sensor and the sensitivity of the sensor integrated over all wavelengths. Knowledge of the sensor's response function is important for colorimetry and the research area of color constancy. Such data may not always be available from the manufacturer of the camera. The sensitivity of the imaging device is a result of the hardware properties of theimaging chip, the lens and filters used, and the post-processing done by the processor contained inside the camera. We will be using an evolution strategy to obtain the sensor response curves of a camera given a single image of a calibration target.

The human visual system is able to perceive colors as approximately constant. This ability is known as color constancy. In contrast, the colors measured by a sensor vary with the type of illuminant used. Color constancy is very important for digital photography and automatic color-based object recognition. In digital photography, this ability is known under the name automatic white balance. A number of algorithms have been developed for color constancy. We review two well-known color constancy algorithms, the gray world assumption and the Retinex algorithm and show how a color constancy algorithm may be integrated into the JPEG2000 framework. Since computer images are usually stored in compressed form anyway, little overhead is required to add color constancy into the processing pipeline.

Evolutionary algorithms apply the processes of variation, reproduction and selection to look for an individual that is capable of solving the task in hand. In order to improve the evolvability of a population, we propose to copy important characteristics of nature's search space. Desired characteristics for a genotype-phenotype mapping are described, and several highly redundant genotype-phenotype mappings are analyzed in the context of a population-based search. We show that evolvability is influenced by the existence of neutral networks in the genotype space. The extent of the neutral networks affects the interconnectivity of the search space and thereby affects evolvability. Species evolving on a non-redundant mapping reach a state of stasis after a few generations; in effect, evolution comes to a halt. However, species evolving on a genotype-phenotype mapping with extensive neutral networks are continuously able to find adaptive mutations and are able to locate higher optima. The existence of highly intertwined neutral networks increases the evolvability of a population.

In order to use co-evolutionary techniques successfully one needs to investigate the dynamics of co-evolution. Assuming an open-ended evolutionary process, it would be desirable to establish the necessary and sufficient conditions which lead to a kind of arms race where species continually adapt in response to one another. In this paper we present a model of competitive co-evolution which is intended to investigate these conditions. In our model, co-evolving species are modeled as points which are placed randomly on a uniform landscape which is deformed by the species. The impact a species induces on its surrounding is not immediate. Instead, the deformation follows the species after some latency period. Evolution is modeled as a simple hill climbing process of the species. We investigate different conditions and their impact on the evolutionary dynamics. Some lead to stasis, some lead to cyclic behavior and others lead to an arms race.

The neutral theory of evolution suggests that most mutations do not cause a phenotypic change. In this case the mapping from genotype to phenotype contains redundancy such that many mutations do not have an appreciable effect on the phenotype. This can result in neutral networks; sets of genotypes connected by single point mutations that map to the same phenotype. A population is able to drift along these networks, eventually encountering phenotypes of higher fitness, thus reducing the chance of becoming trapped in sub-optimal regions of genotype space. In this paper we explore the use and benefit of redundant mappings for evolutionary search. We investigate the properties of several genotype-phenotype mappings by performing random walks along the neutral networks in their genotype spaces. The properties are explored further by performing adaptive walks in which a concept of fitness is introduced. A mapping based on a random Boolean network was found to have particularly interesting properties in both cases.

The size of the search space has been analyzed for genetic programming and genetic algorithms. It is highly unlikely to find any single individual in this huge search space. However, genetic programming with variable length structures differs from standard genetic algorithms where fixed size bit strings are used in that usually many different individuals show the same pheno-typical behavior due to introns. Therefore, finding any given behavior is not as difficult as the size of the search space suggests. A quantitative analysis is presented for the number of individuals that code for the identity function. The identity function is important in the analysis of the search space because it can be used to construct individuals showing the same behavior as any given individual. Finally, an analogy is drawn to nature's sequence space which suggests possible directions for future research. The representation should be chosen such that all possible behaviors are reachable within a comparatively small number of steps from any given behavior and the individuals coding for any given behavior should be distributed randomly in the search space. In addition, long paths of neutral mutations should lead to individuals which code for the same behavior.

An artificial 15 degrees of mobility direct drive hand, slightly bigger than a human hand, is presented. Advantages and limitations of a direct drive hand are given. The authors introduce the design of their hand and present a model for control of the hand. Finally the authors describe their experiments with the hand. The direct drive hand dynamics have been simulated and the control algorithm has been tested on the model.