Repository navigation
预热:在 DVD-R 数据面上刻出彩色角色图 / Teaser: a colour picture burned into the data side of a DVD-R #345
Ashley-QwQ
started this conversation in
展示台 Show and tell
Replies: 1 comment 1 reply
|
A small correction — random data at the beginning didn't help fine pattern picture, my Ritek disc writes still failed with random data at the beginning but Verbatim burned properly even without it. I don't certainly know if it's disc issue, drive issue or both. |
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Click here for English ver ↓
先放图。这是一张普通 DVD-R,没有印刷,也没有涂层。图案是刻进数据面里的:手机打开闪光灯,在盘上方约 21 cm 处往下拍,两个角色就带着颜色显现出来。
颜色不是染上去的。每一小格刻的数据纹理不同,在闪光灯下把不同颜色的光反射进镜头。所以颜色只在设计好的位置成立:手机挪开一些,或者把盘转一下,颜色就会跟着变。
这张盘也是一次测试。右上角那几排小色块是一把"尺子":每排两块,按两种不同的颜色参数刻,拍下来看哪块更接近纯红、纯绿、纯蓝,就能判断参数对不对。结果是两块里有一块明显更接近,说明修正的方向对了,但还多修了一点。
目前颜色模型还不够稳定:有些地方的颜色已经对上了,有些还偏得比较多。所以这部分算法和数据暂时不公开,等稳定以后再放出来。在那之前,先给大家看看现在走到哪一步了。
感谢 @LTVA1。他在 #108 里用显微镜实测了 DVD 上能刻出多细的图案,也是他提出把沿轨采样改成逐字节,disc-probe 后来按这个思路做了更细的网格;他在 GU70N 上反复试刻,发现图案切换太频繁、持续太长时容易刻录失败;细图案在 Ritek 盘上前面垫随机数据也没用,换 Verbatim 盘不垫也能刻过,是盘还是光驱的问题还不确定(感谢他的指正)。另外他在 DIS #331 建议用平板扫描仪做精细标定。
技术细节
角色参考自 PV【官方投稿】メズマライザー / 初音ミク・重音テトSV:https://b23.tv/LIFkHqH 。盘上的画是参考 PV 里的角色重新画的。
English
Picture first. This is a plain DVD-R with no printing and no coating. The picture is burned into the data side: with the phone's flash on, shot from about 21 cm above the disc, the two characters show up in colour.
The colour is not ink. Each small cell carries a different data pattern, and under the flash each pattern sends a different colour of light into the lens. So the colours only hold at the viewing position they were designed for: move the phone a little, or turn the disc, and they shift.
This disc is also a test. The small blocks at the upper right are a ruler: each row has two blocks burned with two different colour settings, and the photo shows which one comes closer to pure red, green and blue. One block in each row wins clearly, so the correction goes the right way, but it went a little too far.
The colour model is not stable yet: some areas already match, others are still off. So this part, the algorithm and its data, stays private for now and will be released once it is stable. Until then, here is where it stands.
Thanks to @LTVA1. In #108 he measured under a microscope how fine a pattern a DVD can actually hold, and suggested sampling the picture per byte along the track, which led to disc-probe's finer grids. His repeated test burns on a GU70N showed that burns tend to fail when the pattern switches very often for a long stretch; for fine patterns, random data at the start did not help on Ritek discs, while Verbatim discs burned fine without it, and whether the disc or the drive is to blame is still open (thanks to him for the correction). He also suggested a flatbed scanner for fine calibration in DIS #331.
Technical details
Characters after the PV 【官方投稿】メズマライザー / 初音ミク・重音テトSV: https://b23.tv/LIFkHqH . The drawing on the disc is a redraw of its characters.
All reactions