kasumimudel

The kasumimudel, often translated as the "haze model" or "fog model," is a conceptual framework used in certain fields, particularly in image processing and computer vision, to describe and mitigate the effects of atmospheric conditions like haze or fog on visual data. It posits that the observed image is a combination of a clear-sky image and an additive scattering component, which represents the light diffused by particles in the atmosphere.

Mathematically, the model typically expresses the observed intensity I(x, y) at a pixel (x, y) as a function of the scene radiance L(x, y), the atmospheric scattering coefficient A, and a transmission map t(x, y). A simplified representation is I(x, y) = L(x, y) * t(x, y) + A * (1 - t(x, y)). The transmission map t(x, y) indicates how much light from the scene reaches the sensor without scattering, and it is often inversely related to the density of haze or fog. The term A represents the atmospheric light, which is assumed to be constant over the scene.

The primary application of the kasumimudel is in image dehazing algorithms. By estimating the parameters of the model, specifically the transmission map and the atmospheric light, it becomes possible to reconstruct an approximation of the original clear-sky image. This is crucial for tasks that require accurate visual information, such as autonomous driving, surveillance, and remote sensing, where atmospheric conditions can significantly degrade image quality and hinder analysis. Different methods exist for estimating these parameters, ranging from statistical approaches based on image properties to more complex machine learning techniques.