Region of interest extraction using combined segmentation in Multispectral Palm Image

Image Acquisition is the method of capturing the image. Entire image is not needed for Enhancement Restoration and Object Recoginition. Image segemention is to get the information which is needed for Authentication System. Processing the segemented image will be cost effective and time effective. Each person can be differentiated based on the physiological and behavioural characteristics. Example for physiological characteristics are palm, fingerprint, iris etc. Palmprint recognition system consists of capturing image, segmentation, feature extraction, matching and result. For any image the noise should be removed and only the region which is needed should be extracted. In Palm Acquisition, the whole palm is captured but it is not required for recognition or authentication. Only certain features such as palmline and texture are extracted from whole palm, which contains the necessary information. In this paper we proposed a methodology for extracting Region of Interest which consists of Otsu thresholding scheme[18], Morphological operations, Canny edge detection, marking reference point and palm image alignment. Experimental results show that the approach is efficient both in computational cost and segmentation quality of Multispectral palm Images. The performance of Region of Interest (ROI) segementation is tested with CASIA MultiSpectral Palmprint Image Database V1.0. ROI extraction for peg free palm image depends on the hand position, there is no constant time. In our methodology ROI is extracted in 2.45 seconds.