PCA Based Generalized Interpolation for Image Super-Resolution

In this paper we propose an eigenimage based superresolution reconstruction technique. Eigenimages of a database of several similar low resolution images are obtained and the given low resolution image is projected on to the eigenimages to compute the eigenimage coefficients. The eigenimages are then interpolated using any conventional interpolation method and approximated to the nearest orthonormal bases. The high resolution image is reconstructed using these bases and the same coefficients determined as above. This method is applicable to images of a particular class and results are demonstrated for both face and fingerprint images. The algorithm offers significant advantage when the input image is blurred and noisy.

• Type: Conference
• Year: 2004
• Where: ICVGIP
• Authors: C. V. Jiji, Subhasis Chaudhuri
• Keywords: ICVGIP 2004, ICVGIP 2008, Low Resolution Image, Resolution Image, Similar Low Resolution