### Table 1. Results of fractal image compression using self-organizing domain classification (method SO ), domain comparison using features only ( FO ), and the baseline fractal image compression method ( Base ).

"... In PAGE 3: ... 5. Results Table1 compares fractal image compression results using three different methods applied to the images shown in Figures 1 and 2. The baseline ( Base ) method is the standard quadtree method as discussed in [2], with no domain classification.... ..."

### Table 1. Results of fractal image compression using self-organizing domain classification (method SO ), domain comparison using features only ( FO ), and the baseline fractal image compression method ( Base ).

"... In PAGE 3: ... 5. Results Table1 compares fractal image compression results using three different methods applied to the images shown in Figures 1 and 2. The baseline ( Base ) method is the standard quadtree method as discussed in [2], with no domain classification.... ..."

### Table 1: Reconstruction fidelity with local search for the fractal scheme based on region-growing ([12]). The image partition con- sists of 1000 range blocks.

2000

"... In PAGE 2: ... When two adjacent (parent) range blocks are merged, a set of candidate domain blocks is formed by appropriately extending the size of each of the domain blocks corresponding to the parent range blocks and retaining the ones that yield the smallest local collage error. Table1 gives the PSNR performance of the code in [12] and that of the local search al- gorithm after trials (third column) and after convergence (last column) for three test images. In Step 2 of the algorithm, for each range block, the search for a new domain block was re- stricted to the local domain pool consisting of the final set of domain blocks.... ..."

Cited by 1

### Table 2: Convergence of the decoding for the three methods for the 512 512 Baboon image. Scheme 3 is based on the pixel frequencies in the fractal code.

1996

Cited by 6

### Table 2: Performance of quadtree type fractal image compression with and without isometries.

1996

Cited by 8

### Table 1: Satellite image compression with individual bands fractal-coded

### Table 1. Robustness of IA-DCT to JPEG compression for garden image

1999

"... In PAGE 7: ... This image has been JPEG compressed to different data rates. Table1 shows the watermark detection statistics for IA-DCT at 1.... ..."

Cited by 97

### Table 3: Image Compression Schemes. Compression

"... In PAGE 12: ... Thus some form of compression schemes are always used that cash on the property of high spatial redundancy in digital images. Some popular compression schemes [32] are illustrated in Table3 . Most modern image compression schemes are progressive, which have important implications to transmission over the communication networks [20].... ..."

### Table 1: Performance of the hybrid scheme for the 512 512 Boat image for di erent settings of the tolerance bounds and . Here N1 is the number of VQ encoded range blocks. The fourth column indicates the compression ratio. The extra bit per range specifying the encoding way (VQ or fractal) is taken into account. Lempel-Ziv coding (Gzip) of the corresponding bit plane is used. The last two columns show the PSNR in dB and encoding time in seconds. In comparison, the distance based fractal code necessitated 57 seconds for the encoding, had a compression ratio of 4.74 and provided a PSNR of 36.29 dB. A full search (256-class search) would increase the PSNR to 36.52 dB.

1996

"... In PAGE 2: ... set. All coe cients were uniformly quantized. The max- imum value of the scaling factor was set to 1 for the fractal part. Table1 shows the results obtained when varying the tolerance bounds and for the 512 512 Boat image. The PSNR was computed without postprocessing.... ..."

Cited by 6

### Table 1: Performance of fractal image compression with and without isometries. The partitioning of the images are of xed range block size.

1996

"... In PAGE 2: ...ere computed using full search, i.e., all domains (with all of their isometric versions, if needed) are checked. The results for the well known test images Lenna, Peppers, and Mandrill are given in Table1 . The im- ages are encoded 18 times with di ering parameters: the domain pool size is large, medium, or small; the partitioning is made from blocks of size 4 4, 8 8, or 16 16, and the encodings use the codebook generated from the domain pool with isometries or a codebook of the same size obtained from a plain domain pool.... ..."

Cited by 8