Iris Image Normalization Method to Pupil Detection with Intensity Transformation
Dewi Nurdiyah, Indra Abdam Muwakhid · 2018
Iris image taken by infrared camera is often applied to security system, and it is called an iris recognation system. On the iris identification system, iris detection is a crucial factor to determine the iris localization. So, if the pupil detection fallacy occurs it can be said that the iris localization is wrong. Every iris image from the acquisition of infrared camera has a different variation of contrast, illumination, and noise so the image normalization stage is need to be done. This research aims to propose the normalization method to detect pupil by changing the image intensity to the new intensity so that the image contrast can be stretched. Therefore, the pupil area is dark and other area is bright. This normalization also aims to maximize the image illumination so that the contrast between dark and bright on the image after contrast stretching can be seen clearer. The method used to this normalization are contrast stretching and gamma which the value gained from the maximize limit of contrast stretching. A very irritating image noise in the pupil detection process is reflection which comes from either inside or outside the pupil area. This noise can be deleted by the closing morphology operation. Then, the result of normalization image is become as an input image to pupil detection with Circular Hough Transform. The result of this experiment from 150 dataset images taken randomly from Biometric Ideal Test 4.0 version using the proposing method shows that an optimal pupil detection accuracy is 99,33%. Therefore, the pupil detection with the proposed method is suitable to be applied to iris recognition system.