Linear filtering for reducing blocking effects in orthogonal transform image coding

Abstract

Working on an image coding method using the discrete cosine transform with 8 × 8 pixel blocks, blocking effects are found. Filtering is used after image reconstruction without modifying the coding method. We have studied some different linear filters, varying the number of neighboring pixels and varying the value of the coefficients of each point. We compare, in each case, Laplacian filters with other filters, and we show the blocking effects and then the filtering effects with EARTH016 and EARTH125 images. Finally, we plot the curve of the mean square error of the reconstructed image depending on the coefficient values used with the One Neighbor Accounting Filter. Various tests have been done with varying images, compression ratios, and MSEs. In conclusion, we compare each filter in each case to determine the best.