A new colour constancy algorithm based on automatic determination of gray framework parameters using neural network

Colour constancy is defined as the ability to estimate the actual colours of objects in an acquired image disregarding the colour of scene illuminant. Despite large variety of existing methods, no colour constancy algorithm can be considered as universal. Among the methods, the gray framework is one of the best-known and most used approaches. This framework has some parameters that should be set with appropriate values to achieve the best performance for each image. In this article, we propose a neural network-based algorithm that aims to automatically determine the best value of gray framework parameters for each image. It is a multi-level approach that estimates the optimal values for the gray framework parameters based on relevant features extracted from the input image. Experimental results on two popular colour constancy datasets show an acceptable improvement over state-of-the-art methods.     

people and personalities

The problem of imposing an in-line velocity perturbation on a uniform flow about a fixed cylinder is studied numerically. A finite volume method using SIMPLEC algorithm is applied on an O-type body-fitted co-located grid. The effect of frequency, amplitude, and shape of the velocity perturbation on the lift and drag coefficients and the Nusselt and Strouhal numbers were studied for the case of Reynolds number equal to 1000. It is noticed that lock-on is present and have a pronounced effect on the average Nusselt number, increasing it by 8.3 percent relative to the uniform flow value, though with a lower effect on drag and lift coefficients. In addition, it is concluded that a sinusoidal perturbation with a dimensionless frequency equal to 0.15 and dimensionless velocity perturbation equal to 0.2 is a better choice to enhance the heat transfer rate.     

A MODIFIED SOLUTION-DIFFUSION MODEL FOR SEPARATION OF ETHANOL-WATER AZEOTROPIC MIXTURES IN PERVAPORATION

A modified solution-diffusion model is presented which describes separation of ethanol-water mixtures. The model is particularly useful in predicting the solute (ethanol) flux as the mixture approaches the azeotrope point where ethanol activity is greater than 0 6. Based on the experimental observations obtained from the literature, the model assumes that polymers of about seven molecules of ethanol are formed, which are then transported through the membrane along with water and ethanol monomers and dimers. It is shown that the ethanol flux can be predicted more accurately as compared with the existing model. Comparisons with experimental data from the literature confirms this model.