Gallery of Before and After Images

This is document presents some of the results obtained from the quadtree algorithm.

babyface Image

Cell 1: Original image babyface without compression.
File size 43,844 bytes.		 Cell 2: Decoded babyface from compressed image file with an RMS tolerance of 1.0 resulting in 3.6:1compression ratio and 12,271 byte compressed file.
Cell 3: Decoded babyface from compressed image file with an RMS tolerance of 8.0 resulting in 12.5:1 compression ratio and 3,498 byte compressed file. Cell 4: Same as picture to left except transformations decoded 5 times rather than 10 and no smoothing.

ww2ships Image

Cell 1: Original image ww2ships without compression.
File size 45,023 bytes.		 Cell 2: Decoded ww2ships from compressed image file with an RMS tolerance of 20.0 resulting in 9.1:1compression ratio and 4,492 byte compressed file.
Cell 3: Decoded ww2ships from compressed image file with an RMS tolerance of 0.5 resulting in 3.5:1 compression ratio and 12,837 byte compressed file. Cell 4: A (medium quality) JPEG compressed version of ww2ships that requires 13,338 bytes of disk space

The original ww2ships image can be thought of as having three main components: the sky, the battleships and the ocean. Encoding the sky is very easy since it is almost uniform throughout and contains no distinguishing features. The battleships also seem to have a self-similar nature about them. The water, conversely, is non-uniform (mostly because of the foam and small waves). Notice that in the second cell (using an RMS an RMS tolerance of 20.0) how blocky the water looks (especially on the left side of the image) while the sky and battleships have fairly good results. The third cell uses very little compression -- that is, the RMS an RMS tolerance is very low and the resulting image has fairly good quality. The last cell uses medium quality JPEG compression and results in a file size similar to that of the encoded (or rather, compressed) image file in the third cell. Clearly, the JPEG compressed image is of a much higher quality.

clouds Image

Cell 1: Original image clouds without compression.
File size 36,447 bytes.		 Cell 2: Decoded clouds from compressed image file with an RMS tolerance of 50.0 resulting in 51.6:1compression ratio and 706 byte compressed file.
Cell 3: Decoded clouds from compressed image file with an RMS tolerance of 8.0 resulting in 26.1:1 compression ratio and 1,396 byte compressed file. Cell 4: Decoded clouds from compressed image file with an RMS tolerance of 1.0 resulting in 2.8:1 compression ratio and 12,956 byte compressed file.

Moving from cells two to three to four we see that the quality of the decoded images increases at the cost of encoding time and disk space. The last and first images are almost indistinguishable -- the compressed image file is almost one-third the file size of the original image.

barnowl Image

Cell 1: Original image barnowl without compression.
File size 36,021 bytes.		 Cell 2: Decoded barnowl from compressed image file with an RMS tolerance of 8.0 resulting in 5.5:1compression ratio and 6,599 byte compressed file. Decoded to 10 iterations.
Cell 3: Same as cell 2 but decoded to 5 iterations. Cell 4: Same as cell 3 but without smoothing.

studentfieldtrip Image

Cell 1: Original image studentfieldtrip without compression.
File size 64,349 bytes.		 Cell 2: Decoded studentfieldtrip from compressed image file with an RMS tolerance of 8.0 resulting in 5.0:1compression ratio and 12,936 byte compressed file.
Cell 3: Decoded studentfieldtrip from compressed image file with an RMS tolerance of 20.0 resulting in 6.2.1:1 compression ratio and 10,379 byte compressed file. Cell 4: Decoded studentfieldtrip from compressed image file with an RMS tolerance of 40.0 resulting in 28.9:1 compression ratio and 2,230 byte compressed file.

 

 

 

Send questions and comments to Farhad Sadaghiani (g9sada@cdf.utoronto.ca)