A System for Automatic Mood-Transferring between Color Images
自動影像心情轉換系統
Abstract
The popularity of digital cameras has brought us an enormous amount
of digital images, thus a huge requirement on processing them. In
particular, the functionality of changing the“tones”or“moods”of an
image is increasingly desired. For example, one may want to convert
the greenish trees into yellowish ones within an image to hint a
seasonal change.
Reinhard et al.s' ground-breaking work of“color transfer between
images”, in many cases, could provide a successful means to satisfy
such a need. However, there are still two issues that remain
relatively unexplored. First, the effectiveness of Reinhard et al.s'
work depends heavily on the provided reference image, but how to
choose an appropriate one is left to users. Second, it is often
difficult to evaluate or quantify the effectiveness of a color
transfer, especially when the input image and reference image
present significant content difference.
In this paper, we tackle the first issue by providing a more
intuitive and user-friendly interface, through which the conversion
of a given input image into a desired style could be done with
simply a mouse click, while without the requirement of a reference
image. To address the second issue, our system also offers a
fine-tune mechanism that allows a more accurate color style
conversion of the input image by employing a“histogram
matching”approach.
摘要
影像處理的需求隨著數位影像的普及日漸增大，其中影像色彩的改變常見於影像處理的領域中，但手動調整影像色彩的過程繁複且費時，因此Reinhard等人提出自動轉換影像色彩的概念，使用者只需提供一張參考影像，做為色調的參考，系統即可自動地將影像的色彩經轉換後具有參考影像的色彩風格，讓使用者在不需手動的情況下即可調整影像的色彩。
在色彩心理學上來說不同的色彩可產生不同的心理作用，而不同的影像色彩也可讓觀看者產生不同的心情，因此我們可對影像做心情分類並結合影像色彩轉換，並依此建構一個依心情分類的影像資料庫，讓使用者可依心情來轉換影像的色彩，在輸入影像並選擇想轉換之心情後，自動將轉換後的結果影像輸出。
為了讓影像轉換的結果能夠在完全符合參考影像的情況下還能保持區域之空間連貫性(spatial coherence)，我們引用了區域分割的概念，提出區域調色與長條圖匹配等兩種方式做色彩微調，使得結果能完全與使用者所選擇之心情一致。