Scale and rotation-invariant feature extraction for color images of iris melanoma

摘要

This paper proposes a feature extraction approach inspired by the Scale Invariant Feature Transform (SIFT) algorithm. We compute the color gradient based on Principal Component Analysis (PCA). To discover the correct keypoints position, we use the hue information of the image and apply our own color corner detection algorithm, described further in this article. We present our results, both on images of human eye iris affected by melanoma and also on images representing color textures of healthy iris. Then we conclude this paper.