1.0 理论基础
简言之
- 图像主要噪声是高斯噪声和脉冲噪声(椒盐噪声)
- sensor噪声特性由信号强度相关与增益相关两方面决定
- 高斯滤波系数来自高斯滤波核,滤波核与sensor噪声特性相关
- 高斯滤波去除噪声时需要考虑局部区域是倾向边缘还是噪声
- 脉冲噪声区域需剔除原始像素点(仅取高斯滤波结果作为最后去噪结果)
2.0 算法过程
2.1 作用位置(ISP pipeline)
![](https://img.haomeiwen.com/i2004776/fcb7c59711df7103.png)
2.2 噪声估计
![](https://img.haomeiwen.com/i2004776/3c4454b11bd7a821.png)
![](https://img.haomeiwen.com/i2004776/5b949876a850c663.png)
3.0 实现思路
![](https://img.haomeiwen.com/i2004776/8c64e7c2599f6138.png)
![](https://img.haomeiwen.com/i2004776/7193b703f2fa6f07.png)
![](https://img.haomeiwen.com/i2004776/203303089930dac7.png)
简言之
本文标题:【论文解读】SS的bayer noise reduction
本文链接:https://www.haomeiwen.com/subject/rklhbctx.html
网友评论