An Enhanced Image Dehazing Method for the Application in Urban Computing

The appearance of images captured in some weather conditions is often degraded due to the presence of haze or fog. These hazy images affect the visibility in the field of computer vision applications, object recognition systems, intelligent transportation, traffic analysis etc. Multi-scale Gradient enhancement is the most popular method used as part of dehazing. But manipulating the image easily turns low dynamic range into high dynamic range images. As a result the restored image become dark or overexposure so that it degrades the quality of the image. As an aim to resolve this problem a better image enhancement as well as classification method is employed. The projected work takes benefit of SVD as this solo value allow us to characterize the picture with a minimal set of value that shrink storage space and progress the quality. For a superior quality and feature improvement a rate based alteration is completed with the help of gaussian filter. The eventual dehazed image is then classified using an improved KNN as part of geographical data analysis. Overall the platform research theme focuses on two key areas-Image processing and data mining.