Accelerated Fog Removal from Real Images for Car Detection
Rawan Younis, Nabil Bastaki · 2017 9th IEEE-GCC Conference and Exhibition (GCCCE) · 2017
Image dehazing improves the visual quality of images in computer vision applications, such as object detection and object tracking. An accelerated image enhancement technique is presented for car detection as part of an effort to count cars using existing street cameras for the purpose of traffic management. Two aspects of car detection are tackled: 1) An existing image fog removal technique is accelerated by replacing a time consuming image filter with a faster filter while maintaining negligible image degradation, 2) A quick and practical algorithm to detect a car in a fog-free image is proposed and applied to a database of about 100 car images. Acceleration is the main goal of this research, in addition to car detection accuracy. The improved fog removal technique is performed by estimating the transmission map using the Proposed Adaptive Filter (PAF) to recover the scene depth of the foggy image. After filtering, a simple, yet exact and effective, car detection algorithm is executed to confirm the presence or absence of a car in the processed image. The system is fairly robust and although all images were obtained from existing sources, the proposed algorithm is expected to perform equally well with any side-view image of a car in the presence of heavy fog and under real conditions.