31-01-2013, 04:35 PM
Fast Multi-exposure Image Fusion with Median Filter and Recursive Filter
1Fast Multi-exposure Image.pdf (Size: 1.8 MB / Downloads: 38)
Abstract
This paper proposes a weighted sum based
multi-exposure image fusion method which consists of two
main steps: three image features composed of local contrast,
brightness and color dissimilarity are first measured to
estimate the weight maps refined by recursive filtering. Then,
the fused image is constructed by weighted sum of source
images. The main advantage of the proposed method lies in a
recursive filter based weight map refinement step which is
able to obtain accurate weight maps for image fusion.
Another advantage is that a novel histogram equalization and
median filter based motion detection method is proposed for
fusing multi-exposure images in dynamic scenes which
contain motion objects. Furthermore, the proposed method is
quite fast and thus can be directly used for most consumer
cameras. Experimental results demonstrate the superiority of
the proposed method in terms of subjective and objective
evaluation.
INTRODUCTION
Images taken by ordinary digital cameras usually suffer
from a lack of details in the under-exposed and over-exposed
areas if the camera has a low or high exposure setting. High
dynamic range (HDR) imaging solves this problem by taking
multiple images at different exposure levels and merging them
together. This technique has been widely used in digital
camera and mobile phone devices. Generally speaking,
existing HDR imaging approaches can be divided into two
categories: tone mapping based methods and image fusion
based methods.
Tone mapping based methods consist of two main steps:
HDR image construction and tone mapping. Multiple low
dynamic range (LDR) photographs are first captured and
combined together to construct a HDR image [2]. Then,
through using tone mapping techniques [3], the overall
contrast of the HDR image is reduced to facilitate display of
HDR images on devices with lower dynamic range. This twophase
workflow i.e., HDR image construction and tone
mapping, can generate a tone mapped image where all areas
appear well exposed. Many effective tone mapping methods
have been proposed [4]-[6].
FUSION OF MULTI-EXPOSURE IMAGES WITH MEDIAN
FILTER AND RECURSIVE FILTER
Fig. 1 shows the schematic diagram of the proposed
weighted sum based image fusion method. The weights of
pixels of different source images are first estimated, and then
refined by recursive filtering with the corresponding source
image serving as the reference image. Finally, the fused image
is constructed by weighted sum of source images.
Color Dissimilarity
When fusing images in dynamic scenes which contain
motion objects, as well as the local contrast and brightness
feature, the influence of motion objects should also be
considered for weight estimation. Through measuring the
color dissimilarity between pixels of source images and pixels
of the scene’s static background, a novel histogram
equalization and median filter based motion detection method
is proposed. Fig. 2 shows the schematic diagram of the
proposed motion detection method.
Weight Estimation
In order to preserve image details and remove influences of
under-exposed pixels, over-exposed pixels, and pixels from
motion objects, the three image features i.e., local contrast,
brightness, and color dissimilarity should be combined
together for weight estimation. The straightforward way to
this objective is by multiplication. However, the pixels of the
same location of different LDR images may be all labeled as
under-exposed, over-exposed or motion objects, and this is
unreasonable especially when these pixels appear in a large
number.
CONCLUSIONS
In this paper, a fast and effective multi-exposure image
fusion approach is proposed. A novel histogram equalization
and median filter based motion detection method is developed
for fusion of multi-exposure images in dynamic scenes.
Furthermore, a recursive filter based weight map refinement
method making full use of the color consistency between
nearby pixels is adopted for weight refinement. Experiments
demonstrate that proposed method can create high visual
quality tone-mapped-like fused images in both dynamic
scenes and static scenes. Furthermore, the effectiveness of the
proposed method is demonstrated by using objective image
fusion quality indexes. The proposed method has been applied
for infrared and visual image fusion. In the future, whether the
proposed method can be applied for fusion of multi-focus
images in dynamic scenes can be further researched.