Fast Efficient Transforms for Contour Extraction from Encrypted Medical Image

Abstract

The necessity of knowing the boundaries of the image is occupies the most important to researchers. With clear conrours, the doctor can easily diagnose the patient's condition. This is possible, but the challenge is whether we can do that for the medical image after it has been encrypted. The encryption algorithm used here is Rivest-Shamir-Adleman (RSA) algorithm which uses two-key encryption, one of them is secret. In this work we introduce a new idea to extract the contours from the encrypted image after converting them to spectral domain methods using Lifting Wavelet, Walsh, and Periodic Haar Piecewise-Linear Transforms. In the specrum image, the compression is done using zonal sampling method. To increase security, the Arnold transform will be applied to the encrypted image using privat keys. The contours extraction from the reconstructed medical image can performed using Canny edge detector. The comparison between those specral algorithms is performed in terms of energy retained, consumed time, compression ratio, and the contour points number which can be detected by the edge detector operator. The experiments results show that by this algorithm, the high number of contour points can be easily detected from the transmitted encrypted medical image and is better using DCT transform for the same compression ratio. However the higher compression ratio using PHL transform is obtained and is exceeds to 88.5391% with retained energy reached to 84.125%.