Severe weather conditions make atmospheric phenomena like haze or fog scatter light and degrade the visual quality of natural images, making them hazy. Hazy images suffer from poor information quality, fainted surfaces, and color shift and may not be suitable for various applications. Haze removal or defogging defined as a technique to reduce or remove interference due to haze. This makes dehazed image highly desired in computer vision and imaging applications. The existing dehazing algorithm majorly lacks accurate estimation scene transmittance, making the dehaze image not suitable in many imaging applications. This talk is aimed at exploring efficient scene transmission algorithm for accurate dehazing results. The method explores convolution neural networks with geometrized pixel difference to optimize transmission map. The algorithm has been tested using natural and synthetic images and results compared with those existing state of the art methods.
Fayadh Alenezi will be speaking at International Congress on Digital Image Processing 2022 which is scheduled to happen on 11th and 12th August 2022 at Hong Kong, HKSAR.