Digital Image Resizing Using Transform Domain Method

Abstract — A simple interpolation-based approach for image resizing is proposed. The Interpolation-based approaches generally suffer from the problem of blurry and jagged edge. Therefore the resizing algorithm is designed to preserve the image content and to minimize the difference between the original image & the resized image. The proposed resizing algorithm can be used for integer as well as non-integer scale factors. It basically consists of 2 operations on the image - Sampling rate change and low pass filtering. The sampling rate change is done by bilinear interpolation. The anti-imaging low pass filter is implemented in 3 ways – 1) FIR low pass filter 2) Wavelet transform and 3) DCT. The performance of the algorithm is tested for 20 different types of images using each method for different scale factors. The test image is scaled by a scaling factor a and a reverse transformation with scaling factor 1/a is carried out using the same method. The loss of information is measured using PSNR. The results of comparisons show that the DCT method outperforms the other two methods. The performance of the algorithm is excellent for the low contrast images.