A RETINEX BASED LOW LIGHT IMAGE ENHANCEMENT METHOD FOR NATURALNESS PRESERVATION

Abstract

Images captured in a poor environment is a big problem to deal with. The environment contains haze, fog, smog, rain or even the lighting conditions could be poor. One of these issues are images captured in low light as they are not fit for the viewability of human beings as well as pose an issue for various computer vision applications, where the image quality needs to high. This research proposes a low light photo improvement method based upon the Retinex theory. The proposed methodology firstly smoothens the image input and with edge preservation characteristic. Then the image is element wise divided by the smooth version of the image. To obtain the Reflectance component. These are then plugged into the structure and texture aware Retinex based equation. This way the low light photo improvement is achieved along with a good amount of naturalness preservation. However, there are pros and cons to this method, which have been discussed in this report.