Mixed-transform image coding with variable block sizes

In the proposed image coding, variable block size and two transforms are us ed in image transform coding. A 16x16 image block is detected to be active or inactive according to activity indexes which are the values of three tra nsform coefficients. If the image block is active, then it is partitioned t o four 8x8 image blocks and then processed by a 2-D 8x8 discrete cosine tra nsform (DCT); otherwise it is processed by a 2-D 16x16 Walsh-Hadmard transf orm (WHT) directly since the WHT has simple structure and the same image co mpression performance as the DCT does for inactive image blocks. as a resul t, the proposed approach is to process image blocks by a 16x16 WHT or an 8x 8 DCT. If the PSNRs for both the conventional DCT image coding and the mixe d-transform coding are kept the same in our experiments with 9 images, the number of additions of the latter increases by 4.18% on average, but the nu mber of the multiplications decreases by 78.53% at bit rate 0.5.