Transformada seno discreta com rotação de eixos bidimensional (DSTr -2D): aplicações na codificação e interpolação de imagens para redução de efeito de blocos

Abstract

This work presents a new tool for image coding and image interpolation based on the two-dimensional discrete sine transform with axis rotation (DSTr-2D). The main characteristics of this transform are the few blocking effects found in the reconstructed image and its computational simplicity. In this work, the DSTr-2D is applied to image interpolation and its performance is compared with results obtained from DCT. A method of interpolation using DCT without extrapolation is also presented. The DSTr-2D achieves a performance better than the other interpolation methods used for comparison purposes. Indeed the better performance is obtained when the sample rate is increased. The DSTr-2D is also applied to the image coding, and it was verified that its advantage relative to DCT in terms of blocking effects in the reconstructed image. The results obtained in image coding using DSTr-2D are compared with results obtained from DCT and LOT. The images were coded using scalar, gaussian, laplacian and Lloid-Max quantization Finally, the DSTr-2D is applied to performs the image coding and the image interpolation, simultaneously.