1. Noise Type
??And unfortunately it's not, actually there's not real physical systems that produce Gaussian noise. So, why is this such an important noise?
- mathematically, very easy to work with
- good approximation to other types of noise, especially for small regions of image or small region of pixel values
??z-是Rayleigh噪聲的均值
??σ是方差
??Rayleigh Noise normally is used to model noise in certain areas of magnetic resonant imaging. This is a good model for real physical devices.
??Uniform是量化噪聲的模型, Exponential是預(yù)測(cè)編碼的噪聲模型
??Noise有時(shí)是從設(shè)備中產(chǎn)生的, 比如傳感器噪聲, 有時(shí)是源于我們對(duì)于圖像的操作, 比如量化
?? with certain probability you change the pixel completely to a new value. And with certain other probability, you change it to a different value. For example, we start that I go over the image and with certain probability I'd change the pixel, let's say to white, and with other probability I change the pixel, let's say to black and that's why it's called salt and pepper. If I changed it to white that's called salt, and if I changed it to black, that's called pepper. 就是在椒和鹽之間來(lái)回跳躍, 所以叫椒鹽噪聲, 椒鹽噪聲對(duì)有些像素影響很大, 而有些完全沒有影響, 當(dāng)它產(chǎn)生影響時(shí), 影響的程度是相同的, 它model的場(chǎng)景是傳感器以一個(gè)很低的概率出故障, 或者某個(gè)像素?zé)龎牧?/p>
2. Noise & Histograms
??原圖的histogram是三個(gè)δ函數(shù), 就是只有三個(gè)peak, 加入噪聲后會(huì)在這三個(gè)peak附近形成與噪聲概率分布函數(shù)相似的形狀(a shape very similar to the actual probability distribution function around each one of the pixel values)
??為什么要講噪聲和直方圖的關(guān)系, 是因?yàn)橥ㄟ^直方圖我們可以來(lái)估計(jì)噪聲的種類和噪聲的參數(shù)
3. Estimating noise
??如果我們知道噪聲的類型, 那我們直接就通過直方圖去算噪聲參數(shù), 然后根據(jù)噪聲類型和參數(shù)來(lái)選取濾波器, 比如Gaussian Noise可能NLM的效果會(huì)比較好, 中值濾波器對(duì)Pepper Noise效果比較好, , if we don`t know the type of noise, 我們的做法就是try, we basically go and fit using standard tools for function fitting, we basically fit the best of each one of the distribution. 那這個(gè)時(shí)候NR就是一個(gè)信號(hào)擬合問題, 用這些標(biāo)準(zhǔn)的噪聲分布函數(shù)來(lái)擬合信號(hào), 哪一個(gè)產(chǎn)生的誤差最小, 我們就選擇哪一個(gè), 有可能不完全是原來(lái)的噪聲分布, 但是我們希望這是一個(gè)對(duì)噪聲的較好近似
4. Degradation Function 退化函數(shù)
?&esmp;完全退化模型, h被稱為模糊函數(shù), 如果知道了退化函數(shù)H, 我們就可以做逆濾波來(lái)重建原始圖像, 如何估計(jì)H就是難點(diǎn)
??這基本就對(duì)調(diào)整和評(píng)估鏡頭有用, 人為放一張只有一個(gè)亮點(diǎn)的圖, 來(lái)估計(jì)模糊的程度
?? motion blur, 估計(jì)退化比估計(jì)噪聲更困難, 因?yàn)槟銢]辦法判斷原圖是什么, 退化操作是什么, 這兩者合在一起了, 你必須通過一些準(zhǔn)則來(lái)區(qū)分開它們, 正如slide中講的, 我只給你5, 你怎么知道是那兩個(gè)數(shù)相加的結(jié)果???
