lambda-image is a python library for processing large-scale 3D spatial and temporal mouse embryo date. the paralleled version is built on Spark.

lambda-image includes the basic module for image processing, like preprocessing, registration, 2-side fusion, segmentation, tracking. the project contains two implementation, a paralleled version and a serial version. the paralleled version is written against Spark's Python API(Pyspark), making use of scipy, numpy, PuLP and pandas

• Python 2.7
• numpy 1.9+
• pandas 0.16.1+
• matplotlib 1.4+
• PyWavelet 0.2.2+
• scikit-image 0.11+
• scipy 0.15+
• pyspark 1.3.0
• Cython 0.21+

The paralleled version is designed to run on a cluster, but currently, I just test it on local mode. Anyway, you can get it work by following steps.

1. first install the requirements by pip

pip install -r requirements.txt

2. compile with cython

make

you can use make clean to clean and make test to launch nosetest.
3. install

python setup.py install

or you can try python setup.py develop and python setup.py develop --uninstall
then you can use import lambda-image

• A main class with methods for initialization of Spark and control of the whole workflow.
• Classes for image processing module,like lambda-image_s_fusion.
• Helper components like lambda-image_s_IO, lambda-image_s_common

Preprocessing:

• stripe_removal(): 去横纹
• intensity_normalization(): 亮度平衡，图像压缩
• saturation(): 饱和度调整
• flip(): 翻转图像
• invert(): 前景/背景转换
• black_tophat(): 黑帽滤波
• subtract_Background(): FIJI减背景
• shrink(): 图像降维
• projection(): 图像栈投影
• smooth(): 图像平滑

Registation&Fusion:

• mutual_information(): 基于互信息的对准
• cross_correlation(): 基于互相关的对准
• execute(): 实施对准向量
• content_fusion(): 基于局部熵的对准
• wavelet_fusion(): 基于小波变换的对准

Segmentation:

• Threshold(): otsu阈值粗分+watershed细分
• Properties(): 计算分割块的属性(坐标,朝向,大小...)
• Clustering(): 将2D分割按距离层次聚类

BSD