file format WebPen - measurements Google today announced a new file format for the images presented, but now that the bad is to replace the aging JPEG. Even with 2 Mbps are full-HD images to load annoying and sometimes just can not be included in a text. There are also Internet access, which will be turned off completely for the same reason images. In poorer countries, divided still often an entire village 56k - there are pictures where you can also see details, then completely annoying to impossible. There is no benefit to such a format it also.
is questionable to me whether the format really brings the needed improvements - would be something to strive for here IMHO ranging from less than 250 kB for Full-HD (it was loaded at \u0026lt;1s).
did this I made a small sample of the photo from the Wikipedia article on
1928er Ford Model A . For the creation of the JPEG files Gimp was used, where there may be differences with other programs. The WebPen files created with the Google program available 'webpconv', said the current approach need to be converted to PNG files back, but this does not influence due to the operation of the PNG format, the quality.
results:
- as it lies there, it has 1.25MB
- saved as PNG, again in Gimp (highest Compression level, this is not the quality, but the computational effort will be affected) is 4MB. This file is used as starting material for further conversions.
- for JPEG files, I use myself typically 85%, resulting in 722 kB.
- than 75% JPEG (since it is then slowly worse, particularly vulnerable is the sign in the background, being built on the green surface clutter), we are at 536 kB.
- WebPen tries themselves, with 82% quality, this is a size of 468 kB. It was striking in comparison to the conversion processes with a given level of quality to me much longer working hours - This is because 'webpconv' determines the ideal image quality over the noise of the image.
- WebPen delivers 100% 1MB
- 75% WebPen still looks good and delivers 384 kB.
- and once crossed the threshold of pain: 50% WebPen. We see losses (at the sign again, this time it is out of focus), but impressive 223 kB.
The reality on the Internet (JPEG 100%) thus yields 1.25 MB. These files are based on the principle "more is better" by the cameras themselves created. There is after all the place does not. JPEG itself you get the whole down to 722 kB (the above mentioned 85%). Underneath each picture starts with the somewhat different Pain threshold on, so you should try manually.
If we apply the "more is better" approach now on WebPen, we are at 1 MB, so schonmal 20% profit. Thanks to the "thinking conversion tools but falling to just not 100% files, but these are already slightly in quality, but very much reduced in size. Thus, you end up with no major experiments at 468 Kb - much smaller than a JPEG would be possible with manual optimization (even the file beyond the threshold of pain is even greater!). The targeted 250 kB are here but not reachable.
Yet another experiment is interesting: The logo of H-Online. Currently, a 3.3-kb Large GIF file. Even as a PNG it shrinks to 2.6 kB. A 85% JPEG is slightly larger at 2.7 kB, wherein whether the extremely simple structure image artifacts occur already here, we increased the file - you see them at normal size, however, since the differences are too small. A somewhat similar optical Egebnis WebPen also provides that 62% will take here. The hit is the file Size 1014 Byte This format is with just under 1 kB! Here, the savings is more than 2 / 3 ...
A more critical conclusion an x264 developer pulls on his blog, what if the competition between them is not so utterly astonished. There you have an image of a full HD movie taken, which provides a 100% JPEG quite astonishing 3.3 MB - far more than the double of my test object. The question was also not "who is less for the same quality?" But the opposite of "who is better at the same size?". Here the desired size of 150 kB can certainly see as more ambitious than, equivalent to see all the resulting images like shit. The clerks are there now, however, believe that the JPEG image "compressed pretty dead" ... At the same file I happen to have tried the opposite approach: webpconv with his automatic promotes here are 85% and 640 kB of days, a JPEG is 85% with 846 kB but much larger.
This apparent contradiction is explained by the different nature of the artifacts. A WebPen is blurred with higher compression - a feature that can also be caused by a bad photographer, so it no direct comparison only notice when it is very strong, but then all the more. JPEG contrast, tends to "funny patterns on smooth surfaces. These can occur in any other way, so you can easily see "this image has been compressed too much." Patterns are stronger, will change nothing to the image perception.
For everyday I think the approach of a desired quality, which is then the amount sought for more than the one where the size is predetermined.
