![](bg1.png)
DEPTH PERCEPTION MODEL EXPLOITING BLURRING CAUSED
BY RANDOM SMALL CAMERA MOTIONS
Norio Tagawa, Yuya Iida and Kan Okubo
Graduate School of System Design, Tokyo Metropolitan University, Hino-shi, Tokyo, Japan
Keywords:
Depth Perception, Shape from Blurring, Stochastic Resonance, Fixational Eye Movement.
Abstract:
The small vibration of the eye ball, which occurs when we fix our gaze on an object, is called “fixational eye
movement.” It has been reported that this vibration may work not only as a fundamental function to preserve
photosensitivity but also as a clue to image analysis, for example contrast enhancement and edge detection.
This mechanism can be interpreted as an instance of stochastic resonance, which is inspired by biology, more
specifically by neuron dynamics. Moreover, researches for a depth recovery method using camera motions
based on an analogy of fixational eye movement are in progress. In this study, using camera motions espe-
cially corresponding to the smallest type of fixational eye movement called “tremor.” We have constructed
the algorithms which are defined as a differential form, i.e. spatio-temporal derivatives of successive two im-
ages are analyzed. However, in these methods, observed noise of derivatives causes serious recovering error.
Therefore, we newly examine a method in which a lot of images captured with the same camera motions are
integrated and the observed local image blurring is analyzed for extracting depth information, and confirm its
effectiveness.
1 INTRODUCTION
Camera vibration noise is serious for a hand-held
camera and for many vision systems mounted on mo-
bile platforms such as planes, cars or mobile robots,
and of course for biological vision systems. The com-
puter vision researchers traditionally considered the
camera vibration as a mere nuisance and developed
variousmechanical stabilizations (Oliverand Quegan,
1998) and filtering techniques (Jazwinski, 1970) to
eliminate the jittering caused by the vibration.
In contrast, a new vision device, called the Dy-
namic Retina (DR), which directly takes advantage
of vibrating noise generated by mobile platforms to
enhance spatial contrast (Propokopowicz and Cooper,
1995). Furthermore, for edge detection, the Resonant
Retina (RR) indicating the DR model with the tech-
nique based on stochastic resonance (SR) is proposed
(Hongler et al., 2003). SR can be viewed as a noise-
induced enhancement of the response of a nonlinear
system to a weak input signal, for example bistable
devices (Gammaitoni et al., 1998) and threshold de-
tectors (Greenwood et al., 1999), and naturally ap-
pears in many neural dynamics processes (Stemmler,
1996).
Although DR and RR offer their massive paral-
lelism and the simplicity of their architecture, by con-
sidering especially the enough potential of the cam-
era vibration for depth perception, we have proposed
shape recovery methods using the camera motion
model imitating fixational eye movements (Tagawa
and Alexandrova, 2010), (Tagawa, 2010). These
methods are constructed based on a differential form,
and the gradient method for ”shape from motion”
(Horn and Schunk, 1981), (Simoncelli, 1999), (Bruhn
and Weickert, 2005) is used fundamentally in order
to recover dense depth map with low computational
cost compared with the methods based on correla-
tion matching. The fixational eye movement is clas-
sified into three types as shown in Fig. 1: microsac-
cade, drift and tremor. Here, we focus on the tremor,
which is the smallest one of the three types, to reduce
the linear approximation error of the gradient equa-
tion. However, in this case, we cannot get enough
information to recover accurate depth from succes-
sive two images. Therefore, we have to collect the
enough information about depth from other sources.
Using a lot of images captured with random small
motions of camera, which consists of 3-D rotations
imitating fixational eye ball motions (Martinez-Conde
et al., 2004), many observations can be used at each
pixel, i.e. many gradient equations can be used to re-
cover the each depth value corresponding to the each
pixel. It should be noted that since the center of the
329
Tagawa N., Iida Y. and Okubo K..
DEPTH PERCEPTION MODEL EXPLOITING BLURRING CAUSED BY RANDOM SMALL CAMERA MOTIONS.
DOI: 10.5220/0003817403290334
In Proceedings of the International Conference on Computer Vision Theory and Applications (VISAPP-2012), pages 329-334
ISBN: 978-989-8565-03-7
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)