多参数融合算法提升自动对焦性能
188 浏览量
更新于2024-08-27
收藏 431KB PDF 举报
本文主要探讨了一种多参数融合算法（Multi-Parameter Fusion Algorithm, MPFA）在自动对焦（Auto-Focus, AF）中的应用。该算法结合了图像锐度评价算法（Image Sharpness Evaluation Algorithm, ISEA）和变焦跟踪方法（Zoom Tracking Method, ZTM）。在AF过程中，算法的关键参数包括变焦电机位置（zoom motor position, z）和背景复杂度（background complexity, c）。通过建立一个根据z和c值确定优先级的表格，算法能够智能地动态调整，即当背景复杂度较高时可能优先采用改进的ZTM，而当变焦电机接近目标位置时则可能切换到改良的ISEA。
在硬件实现方面，论文详细介绍了这种MPFA在德州仪器（Texas Instruments）DaVinci数字信号处理器上的实施。对比传统的AF方法，实验结果显示，该算法能提供显著更快的聚焦速度，特别是在复杂的环境条件下，如高背景复杂度和快速变焦操作时，其性能优势更为明显。
文章的关键词包括图像清晰度、变焦跟踪、优先级表以及与AF相关的特定技术代码000.3110、040.1490、110.2960和110.5200。整体来看，这篇研究对于提升相机系统在实时场景下的对焦效率具有重要意义，是当前光学成像领域中值得关注的技术进步。
580 CHINESE OPTICS LETTERS / Vol. 8, No. 6 / June 10, 2010
Multi-parameter fusion algorithm for auto focus
Jun Luo (ÛÛÛ )
∗
, Li Sun ( ååå), Kesong Wu (ÇÇÇttt),
Weimin Chen (¬¬¬), and Li Fu (GGG www)
Key Laboratory of Optoelectronic Technology and System, Ministry of Education,
Chongqing University, Chongqing 400030, China
∗
E-mail: luojun@cqu.edu.cn
Received January 6, 2010
A multi-parameter fusion algorithm (MPFA) for auto-focus (AF) is discussed. The image sharpness evalu-
ation algorithm (ISEA) and zoom tracking method (ZTM) are combined for AF. The zoom motor position
(z) and background complexity (c) are regarded as the main parameters of this algorithm. A priority table
dep ending on z and c is proposed. Modified ISEA or ZTM is adopted according to the priority table value.
The hardware implementation of the MPFA on Texas Instruments’ Davinci digital signal processor is also
provided. Results show that the proposed scheme provides faster focusing compared with the conventional
approaches.
OCIS co des: 000.3110, 040.1490, 110.2960, 110.5200.
doi: 10.3788/COL20100806.0580.
At present, zoom lenses have been utilized widely in var-
ious industrial applications. Auto-focus (AF) algorithm
is an important factor in determining the imaging qual-
ity of a digital still camera (DSC)
[1−3]
. There are two
ways to implement AF, namely, active AF and passive
AF. In literature, only passive AF methods have been
discussed, although some conventional passive AF meth-
ods have been introduced. In this letter, we introduce a
new AF approach, the multi-parameter fusion algorithm
(MPFA), which merges the image sharpness evaluation
algorithm (ISEA) and the zoom tracking method (ZTM)
according to a priority table. The priority table is based
on zoom motor position (z) and background complexity
(c). Results show that the MPFA provides higher focus-
ing speed compared with conventional methods.
There are two prevalent methods used in realizing
passive AF: ISEA and ZTM
[4]
, and each has many
algorithms, including sharpness evaluation operators,
look-up table, geometric, and adaptive zoom tracking.
However, these algorithms have several general draw-
backs. Firstly, a complex background image results in
local peak values of the image definition. In addition,
more time is required in searching for the maximum
definition value when using ISEA
[5−14]
. Secondly, the
look-up table method uses a large amount of memory,
and no mechanism is provided to select the right trace
curve when the zoom motor is moved towards the tele-
angle direction; this is due to the one-to-many mapping
problem. Thirdly, in the geometric ZTM, offset between
the estimated trace curve and true trace curve gradually
increases for the zoom motor positions towards the tele-
angle direction. Fourthly, the adaptive ZTM improves
the tracking accuracy at the expense of tracking speed
[4]
.
Thus, a more effective AF approach is necessary. The
new MPFA presented in this letter can solve these prob-
lems in various ways.
There are numerous evaluation functions for digital im-
age sharpness, including variance operator, power gradi-
ent operator, and Laplacian operator. The performance
of each operator is different in relation to focusing speed.
The output of high-pixel color charge-coupled
device/complementary metal-oxide semiconductor
(CCD/CMOS) image sensors are RAW format image
data. In conventional methods, data are restored in
BMP or JPEG formats, the color space conversion is
taken, and image definition is evaluated via luminance
or green (G) component. However, these methods are
not so efficient that they require a large quantity of cal-
culation and are also time-consuming.
According to the above discussion, the gradient oper-
ator model, which is directly based on RAW format, is
provided by
F =
X
mat
x
X
mat
y
(f
2
x
(mat
x
, mat
y
) + f
2
y
(mat
x
, mat
y
)), (1)
where
g(mat
x
, mat
y
) = 0.299 × R(mat
x
, mat
y
) + 0.587×
G
0
(mat
x
, mat
y
) + 0.114 × B(mat
x
, mat
y
), (2)
f
x
(mat
x
, mat
y
) = g(mat
x+1
, mat
y
)−
g(mat
x
, mat
y
), (3)
f
y
(mat
x
, mat
y
) = g(mat
x
, mat
y +1
)−
g(mat
x
, mat
y
). (4)
In the above expressions, “mat” is a MAT unit, which
has one red (R) element, two G elements, and one blue
(B) element in the RAW format image acquired. In ad-
dition, (mat
x
, mat
y
) is the MAT unit with coordinate (x,
y), F is the focus value, g(mat
x
, mat
y
) is the gray value
in (mat
x
, mat
y
), and G
0
(mat
x
, mat
y
) is the average of
the two G-components in (mat
x
, mat
y
).
The detailed gradient operator based on RAW format
is not presented here. When obtaining the focus value of
each frame according to Eq. (1), hill-climbing algorithm
is used to find the in-focus position.
ZTM depends on the zoom motor positions and the
focus motor positions conjointly
[15−18]
. Figure 1 illus-
trates the ZTM system. Cam mechanism is adjusted by
zoom motor to change the combination focal length f of
1G and 2G lens units. Zoom PI and focus PI represent
1671-7694/2010/060580-04
c
° 2010 Chinese Optics Letters
下载后可阅读完整内容，剩余3页未读，立即下载
2022-04-23 上传
2017-11-03 上传
点击了解资源详情
点击了解资源详情
点击了解资源详情
点击了解资源详情
点击了解资源详情
点击了解资源详情
点击了解资源详情
weixin_38633967
- 粉丝: 7
- 资源: 930
我的内容管理
展开
最新资源
- 深入浅出：自定义 Grunt 任务的实践指南
- 网络物理突变工具的多点路径规划实现与分析
- multifeed: 实现多作者间的超核心共享与同步技术
- C++商品交易系统实习项目详细要求
- macOS系统Python模块whl包安装教程
- 掌握fullstackJS：构建React框架与快速开发应用
- React-Purify: 实现React组件纯净方法的工具介绍
- deck.js：构建现代HTML演示的JavaScript库
- nunn：现代C++17实现的机器学习库开源项目
- Python安装包 Acquisition-4.12-cp35-cp35m-win_amd64.whl.zip 使用说明
- Amaranthus-tuberculatus基因组分析脚本集
- Ubuntu 12.04下Realtek RTL8821AE驱动的向后移植指南
- 掌握Jest环境下的最新jsdom功能
- CAGI Toolkit：开源Asterisk PBX的AGI应用开发
- MyDropDemo: 体验QGraphicsView的拖放功能
- 远程FPGA平台上的Quartus II17.1 LCD色块闪烁现象解析
