魔法使黎明期 / Mahoutsukai Reimeiki / 魔法使い黎明期 BDRip

原盘画质极具实验性质，向我们充分展示了现代动画工业可以把画面荼毒到什么程度。不知道出于什么艺术意图，正片有极重的彩色动噪点。在如此厚重的负担下，欠码自不必说，搭配 DMM 的劣质编码画质更是雪上加霜，不仅亮度平面质量欠佳，色度平面也有着极佳的研究价值，彩色色块到处都是，像是在看上世纪的彩色显像管电视。正片原生分辨率大概在 845.7p，不过噪点和欠码的干扰太大，我们无法进行准确分析，因此也没有进行逆向拉升重构。全片还有不少 ppt 制作，可能是为了看起来富有“动态"，ppt 分镜附加了大量平移和缩放，可惜在如此糟糕的画质负面加成下，分镜帧间的锐度稳定性也略有欠缺，加重了处理难度。除此以外的瑕疵就显得不那么严重了：欠码造成的蚊噪、色带和振铃，以及较重的锯齿。
The source quality is extremely full of research value, showing us just how far the modern animation industry can poison the image. We don't know what the artistic intent is, the main episodes are extremely heavy with dynamic colour noise. Not to mention the insufficient bit-rate that comes with such a heavy burden, the poor encoding quality with the DMM is compounded by the poor quality of not only the luma plane but also the chroma planes, with colour blocks all over the image, as if you're watching a colour tube TV from the last century. The native resolution is around 845.7p, but due to the very much noise as well as insufficient bit-rate, however, we give up applying rescaling. There's also a fair amount of slide-like production, with a lot of panning and zooming, probably to make it look "dynamic". Unfortunately, with such a terrible image quality, the inter-frame sharpness stability is also slightly poor, making the image more difficult to process. In contrast, other artifact seems less serious: DCT noise due to insufficient bit-rate, colour banding and ringing, and heavy aliasing.

由于是实验性质的项目，处理上相比过往更具激进倾向。首先是彩噪，我们认为本片没有复古的价值，因此使用了神经网络算法在 RGB 空间中进行了码率自适应降噪，随后辅以常规但强度略高的亮度自适应时域降噪。接着就是常规操作：较强的抗锯齿、较强的去色带、常规去振铃、补偿性锐化。最后我们还添加了一些精噪以确保时域稳定。
Since this is an experimental project, the processing is more aggressive than in the past. First, the colour noise, which we thought meaningless, so we performed bit-rate adaptive denoising based on a neural network algorithm in RGB colour space, followed by regular, but slightly more intense, luma-adaptive temporal denoising. This was followed by the usual operations: strong anti-aliasing, strong de-banding, regular de-ringing, and compensatory sharpening. Finally, we added some static noise to ensure temporal stabilization.

由于色度平面质量太差，我们对色度平面有较高强度的处理，部分高鲜艳度区域会有目视损失，不过也很难做到完全分离，权衡之下也就如此处理了。做到一半其实打算弃坑了，发布此片就是主要抛砖引玉，看看有无能人巧匠来挽救一下这朵奇葩。
The quality of the chroma planes are so poor that we have a higher intensity treatment of the chroma planes, with some visual loss in the high vibrance areas, but it's hard to separate them, and the tradeoff is that it's handled that way. Halfway through the process, we plan to give up this project. The main purpose for posting this project is just a throwaway, to see if there is any better scripter who can save this oddball.

感谢所有参与制作者 / Thanks to our participating members:
总监 / Script: 零
压制 / Encode: 莉莉
整理 / Collate: LKey Eden
发布 / Upload: 佣兵
分流 / Seed: VCB-Studio CDN 分流成员

感谢所有资源提供者 / Thank to all resource providers:
CDs: lucasn0tch@jpopsuki

VCB-Studio 已不再保证收集作品相关 CD 和扫图资源，详情请参见 https://vcb-s.com/archives/14331。
Please refer to https://vcb-s.com/archives/14331 for more information about that VCB-Studio will no longer guarantee to include relevant CDs and scans.

本组（不）定期招募新成员。详情请参见 https://vcb-s.com/archives/16986。
Please refer to https://vcb-s.com/archives/16986 about information of our (un)scheduled recruitment.

播放器教程索引： VCB-Studio 播放器推荐
中文字幕分享区： Anime 分享论坛
项目计划与列表： VCB-Studio 项目列表
特殊格式与说明： WebP 扫图说明

Comparison (right click on the image and open it in a new tab to see the full-size one)
Source________________________________________________Encode