A family of methods based on automatic segmentation for estimating digital camera noise: A review

Digital camera noise information is useful in various applications: image enhancement and post-processing, device selection for a particular system, recognition, decision making, and forensics, etc. Therefore, there is a need for fast and reliable noise estimation methods. There are set of methods based on automatic segmentation of targets: uniform (ASUT), non-uniform (original and modified ASNT), strip (ASST) ones and using a single image (ASSI). This paper discusses these methods grouped into one family. The methods in this family estimate noise parameters by processing only a few images. The results are compared with the EMVA 1288 Standard. The simplest and least time consuming methods are ASSI and ASNT. However, they do not allow the estimation of all noise characteristics. ASST and ASUT provide all noise data, but ASST is more complex due to the target design. ASUT can be considered the most balanced and accurate.