Brightness vs. Lightness Enhancement for Image Segmentation

Videos which are taken under low light environment usually contain lots of noise with grainy look. The low light situation likewise yields under-saturated and low contrast video footages. The frames of this kind video footage, if further processed, will be the type of input that are very challenging to deal with. The aim of this work was to choose the best feature candidate, either brightness or lightness, for improving the frames quality which are then fed to the segmentation algorithm. A random residence and a campus parking lot area are the places where the video footages of this research are taken. The first step after gathering the videos is to extract the frames and converting it to 2D images, where the brightness and lightness features of the images are then solely fed into the CLAHE enhancement algorithm. The enhanced images are then fed to the chosen segmentation algorithm, called the Active Contour. From the experiments, it is proven that the enhancement based on the lightness feature outperformed the brightness based enhancement system which is proven by the better segmentation results. Moreover, from the appearance, the output of the lightness based enhancement system looks softer and the resulting image contains less artifacts.