# LESH (Local Energy based Shape Histogram) feature Extraction
# Usage:
lesh_vect = calc_LESH(im) ;
# Input:
im = Image or Local patch
# output:
lesh_vect= LESH feature vector (128 dimentional for 16 partitions and 512 dimentional for 64 partitions of the
image/patch. (see FeatureParam.m)
Works on image or patch of any size (preferably square).
Recomended partition size w is 8 (512-dim vector) for an image of size greater than 64x64. In patches of size 32x32, partition size should be changed to 4 (to yield a 128 dim vector)
There are a number of parametes to adjust for different applications.
file FeatureParam.m may be modified to change e.g. the number of scale and orientation of GABOR filter. partion size may be changed to yield a longer vector or vice versa, more coarse partions or more fine partions may affect the discrimination quality of feature vector.
Recomended settings are 8x8(64) partition size(512-dimenstional vector) and 5 scales and 8 orientations for GABOR filter bank.
# Please Note:
This updated optimized version does not include the Gaussian weighing as described in the paper. With the current optimzation, we have found it works better for general shape description.
Acknowledgemnt: The code uses updated version of phasecongruency detection from Peter kovesi Matlab toolbox.
Please use the respective paper citation in the file phasecong3.m
The software is provided "as is", without warranty of any kind.
# References:
[1] M.S.Sarfraz and O.Hellwich,"Head Pose Estimation in Face Recognition across Pose Scenarios"
in Int. Conference on Computer Vision Theory and Applications VISAPP, Vol. 1, pp. 235 -242, Portugal, 2008.
[2] M.S.Sarfraz, O.Hellwich," An Efficient Front-end Facial Pose Estimation System for Face Recognition",
In International Journal of Pattern Recognition and Image Analysis, distributed by Springer, Vol.18(3) pp. 434441, 2008.
[3] M.Saquib Sarfraz and Olaf Hellwich, "On Head Pose Estimation in Face Recognition",
J. Braz, A. Ranchordas, H. Arajo and J. Jorge Editors, Advances in Computer Graphics and Computer Vision.
Lecture notes CCIS 24, pp 162-175, Springer, 2009.
没有合适的资源?快使用搜索试试~ 我知道了~
人脸图像特征提取matlab代码-LESH:基于局部能量的形状直方图(LESH)特征描述符代码
共7个文件
m:4个
license:1个
md:1个
需积分: 30 1 下载量 52 浏览量
2021-05-22
06:47:51
上传
评论 2
收藏 26KB ZIP 举报
温馨提示
人脸图像特征提取matlab代码LESH(基于局部能量的形状直方图)特征提取 用法: lesh_vect = calc_LESH(im); 输入: im =图片或本地补丁 输出: lesh_vect = LESH特征向量(图像/补丁的16个分区为128维,而64个分区为512维。(请参阅FeatureParam.m) 适用于任何大小(最好是正方形)的图像或补丁。 对于大小大于64x64的图像,建议的分区大小w为8(512像素矢量)。 在大小为32x32的补丁中,分区大小应更改为4(以产生128个暗淡矢量) 有许多参数可以针对不同的应用进行调整。 可以对FeatureParam.m文件进行修改以更改例如GABOR过滤器的比例数和方向。 可以更改部分大小以产生更长的向量,反之亦然,更多的粗糙部分或更多的精细部分可能会影响特征向量的区分质量。 推荐的设置为8x8(64)分区大小(512维矢量),并为GABOR滤波器组设置5个比例和8个方向。 请注意: 此更新的优化版本不包括本文所述的高斯加权。 通过当前的优化,我们发现它对一般形状的描述效果更好。 致谢:该代码使用了Peter kovesi
资源详情
资源评论
资源推荐
收起资源包目录
LESH-master.zip (7个子文件)
LESH-master
phasecong3.m 26KB
LICENSE 34KB
README.md 2KB
calc_LESH.m 3KB
FeatureParam.m 411B
lowpassfilter.m 950B
Read me.txt 3KB
共 7 条
- 1
weixin_38629362
- 粉丝: 6
- 资源: 967
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功
评论0