Analysis of Non-linear Filtering Techniques based on Quantitative Metrics using Different Images

Abstract

Image filter is the process of removing various types of noise from the images. The ensuant image is an information rich image than the original input image. The filtering finds its application in many fields from medical imagery, face detection, robot navigation, object detection, aircraft maintenance to image enhancement and image restoration. In the field of medical sciences the filter serves the purpose of image enhancement for efficient disease diagnosis, in aircraft maintenance for the purpose of detection of faults during takeoff, in case of face detection, object recognition and robot navigation used for object detection. This paper uses different quantitative metrics to analyze the result of different filtering techniques on an image. Initially, well known registered images from various aspects of science and nature are taken such as one image ct.jpg from medical sciences, two images Lighthouse.jpg, Penguins.jpg of natural scenery, two images of faces Koala.jpg, lena.jpg and a picture of naturally grown flowers Tulips.jpg are taken as input. Filtering techniques namely Median Filter (MF), Adaptive Filter (AF), New Adaptive Median Filer (NAMF), New Adaptive Spatial Filter (NASF), Edge Preserving Smooth Filter (EPSF) are applied on them. Further the filtered images are analyzed using five quantitative metrics such as Entropy (EN), Standard Deviation (SD), Peak Signal to Noise Ratio (PSNR), Mean Square Error (MSE), and Mean Absolute Error (MAE). From the experimental result and the corresponding metrics used we observed that the resultant image is more informative than the original source images.