This morning I woke up at 9:45. I naturally wake up at that time when I’m late for work (my alarm is at 8:50). I took half day off (till 1PM) and slept more or less what I needed to.
The night before I went to sleep again at 3AM, plus I went to the gym. Basically very tired !
When I woke up again at 11:30 I found a bad surprise: the router, a Linksys WRT54G, was dead.
I was tempted to go buy a new one after work, but instead I came home and tried fix it following some hints found on the net. Of course it didn’t work. I even opened the thing.. bha. Eventually I settled for using a backup router that I had laying around. Not so easy.. I couldnt find a power adapter that work work for it !
Luckily, I found one that seems to work although it gives 1A instead of specified max of 1.5A. Of course after I did have to spend some time to manage to log-in and setup the Internet connection. I didn’t have the manual, plus the interface is in Japanese.
Now everything seems OK. But eventually I’ll need to buy a new router.. although in the settings I may have fixed the connection timeout problem that I had before with it and that made me buy a new one.

Anyhow, yesterday night, after posting, I went to look for color space infos. I’ve been using YUV (specifically YCrCb) for a long time without quite realizing that it was still playing with RGB color channels. Y is basically a gray produced mostly out of green, Cr a difference mostly in red and Cb same but for blue.
This is the format used for JPEG, MPEG, etc. It comes from the TV standard of encoding image data, but I don’t think it’s the best color space to work for image compression.
A lossy YCrCb signal can bring out some serious visible artifacts if Cr and Cb aren’t compressed equally well.
One has Y (the gray image) that needs high quality and Cr and Cb that need to be compressed so that loss of information happens in a similar fashion. Basically, if with RGB one must be very careful because those are all primary channels. With YCrCb one has more freedom to compress channels separately, but not total freedom, because the Cr and Cb channels must still keep a good correspondence.
This brings me to think that I should try work in a different color space, one that still has 3 channels, but 3 channels that can be compressed independently without bringing out major color artifacts (for the human eye).
The ideal color space should be something that separate the gray scale image component, a basic color palette (a hue) and a saturation for the color. If saturation is bad, the color is still faithful, it doesn’t shift towards red or blue like YCrCb.

Now it’s time to sleep, but I’m happy to know that I have a good direction in which to move now.

zzzzzzzzzzzzzzz

Posted by Davide Pasca in Uncategorized