page 1  (27 pages)
2to next section

26 7. Appendix

7. Appendix

In order to give the reader a flavor of how the JPEG algorithms affect the reconstructed images, we attached three pairs, an original image and its difference image generated from our simulation. In the beginning, we tried to take photographs of the original images and the decoded images. But it turns out that it is difficult to see on the photographs the differences between the original image and the decoded image, although they are visible on the screen. Therefore, instead, we have attached pictures of the difference between the original image and the decoded image. The original images are visually better on the screen than on the photograph due to limitations of the photographic techniques.

Most of the reconstructed images are blurry compared to the original ones. The example luminance quantization table (see Table 3.2) has a better impact on the images Tree, Sailboat, Airplane and Aerial than the images House, Splash, Girl, Lenna and Tiffany. There are visible block edge effects and obvious blurring in the latter decoded images. Since the absolute difference images are too obscure to show on the photographs, we exaggerated these differences in the following way:

ffl For each pixel compute a diff, the absolute value of the difference between the original image and the decoded image.

ffl In a given image, there are very few diffs in the range 128?255. Map those values to pixel value 255. On the screen those will show up as very bright spots, but on the photographs they are not so obvious.

ffl Find the minimum and maximum of all the diffs in a given image.

ffl Exaggerate the differences. If a diff is in the range ?127, then map it to ?255 as
follows:

255 * (diff ? min)/(max ? min)

In the exaggerated difference images it is easy to see the effects of the JPEG algorithms.