SmartCam to see through darkness
Jian Ren Wang, Che Han Su, John Y. Chiang · 2016
The paper proposes a structure for enhancing nighttime image, including the parts of glow removal, nighttime image enhancement and color calibration. Under the prerequisite that nighttime image is commonly affected by intense artificial light (e.g. streetlamp, car light, traffic lights), which violates the existing image processing technique, it is necessary to remove the glow produced by artificial light, and correct the area affected by artificial light. The paper combines negative film technique with dark channel prior (DCP) algorithm to enhance nighttime image, and employs guided image filtering to make transmission more refined, achieving better restoration effect and making speed faster than the conventional soft matting algorithm by several times. Finally, the paper conducts color calibration of the enhanced nighttime image, and calibrates the distorted colors. This algorithm can be used to enhance the nighttime image taken by closed-circuit television (CCTV) or car video recorder, and especially achieves excellent effect when applied in low-light environment.