Contourlet Cryptography: A Better Description for Pattern Recognition

In the application of Computer vision research, Wavelets have enjoyed widespread popularity. In recent perspective some new age transforms like the corex transform, steerab le Pyramid, ridgelets, contourlets and curvelets are theoretically appearing to the better image description for pattern recognition. These new transforms have been applications in the area of compression, denoising, watermarking and digital signature systems. In this work we will compare the performance of wavelets and contourlets for the purpose of pattern recognition cryptography. The results are based on well known database experiments viz USPS database of handwritten numerals and the Essexface database. We chose K-Nearest Neighbor and Probabilistic Neural Network for the purpose of classification. The result indicates that although contourlets surpass wavelets for other image processing tasks like compression and denoising, they are not good as wavelets for the purpose of pattern recognition until or unless cryptography is involved.