[其它] Effective Despeckling of HDR Images

[复制链接]

彬彬

797 主题	1 听众	1万积分

资深设计师

Rank: 7 Rank: 7 Rank: 7

纳金币: 5568
精华: 0

电梯直达

楼主

发表于 2012-1-5 08:59:27 |只看该作者 |倒序浏览

1 Introduction

High Dynamic Range (HDR) images are common in computer

graphics. In these images, speckles can appear for various reasons,

such as high variance for Monte-Carlo-based rendering engines.

These speckles are responsible for large artefacts if nonrobust

post-processing methods are used, such as tonemapping operators

relying on a maximal or average luminance value. Ensuring

that all post-processing operators are robust is tedious, therefore

we propose to handle these speckles before any other processing is

done. This way, we ensure that any HDR image post-processing

pipeline produces acceptable results.

In the specific case of Monte-Carlo-based rendering, sample-space

methods such as [DeCoro et al. 2010] or [Pajot et al. 2011] process

the samples during the rendering process to detect those that

can cause bright spots. These methods pre-suppose a stochastic

rendering method with known properties such as samples independence,

but provide a complete control over the committed error.

Image-based methods are not based on any asumption on the way

the image is produced. In addition of being more general, they require

less computational power and are often easier to integrate in

an image-production pipeline.

A popular image-based method to remove speckles is to use

bilateral-filtering or more specific versions [Xu and Pattanaik

2005], but these methods induce a noticeable blur and artefacts

(Figure 3, 3rd row and Figure 4, 3rd row). Our method is an imagebased

method, but it is designed to not induce any blur nor artefacts.

Our approach consists of two steps: we first detect speckles, and

then recons***ct them using pixels not tagged as speckles. This

is highly different from (bilateral-) filtering methods, which process

all pixels the same way, recons***cting each pixel using all

their neighbours. Our method does not introduce blur, naturally

preserves edges and thin image features, yielding an artefact-free

recons***ction.