A Novel Approach of Image Interpolation using DST
Jagyanseni Panda, Sukadev Meher · 2019
A novel interpolation technique using a Discrete Sine Transform (DST) is presented here. For fractional pixel prediction, DST is used but DST for image interpolation fails to produce pleasant output. The newly developed method implements a solution by further sub-sampling the Low Resolution (LR) image and finds out error by up-sampling it using DST. The error image which contains high frequency variation is responsible for sharpening the interpolated image. So it is added to available LR images. Then, the resultant image is up-sampled by using DST again to get back to the original size. Interpolated image is of high visual quality in the edge and texture part of the image. From the experimental results, it can be demonstrated that this algorithm performs better both objectively and subjectively. Experimental results reveal that it performs Peak Signal to Noise Ratio (PSNR) 2.3436 dB better upon Discrete Cosine Transform (DCT) and 2.3098 dB upon Bi-cubic interpolation.