Nighttime sky/cloud image segmentation
With the spirit of reproducible research, this repository contains all the codes required to produce the results in the manuscript: S. Dev, F. M. Savoy, Y. H. Lee and S. Winkler, Nighttime sky/cloud image segmentation, Proc. IEEE International Conference on Image Processing (ICIP), 2017.
Please cite the above paper if you intend to use whole/part of the code. This code is only for academic and research purposes.
The codes are written in python and MATLAB.
The nighttime image segmentation dataset can be downloaded from this link. A few sample images can be found in the folder
color16Norm.mGenerates the 16 color channels in the form of a MATLAB struct. All values are normalized.
color16_struct.mGenerates the 16 color channels in the form of a MATLAB struct.
createSPImage.mGenerates the quantised- and binary- image of our proposed method.
createSPImageNumber.mGenerates the quantised- and binary- image of our proposed method, based on the number of superpixels.
gacal.mImplements the Gacal approach.
global_th_novi.mImplements the Yang et al. 2009 approach
internal_calibration.pyImplements the internal calibration of our sky camera.
local_th_novi.mImplements the Yang et al. 2010 approach.
RGBPlane.mExtracts the red-, green-, blue- plane of an input image.
score.mCalculates precision, recall, fscore and error of a binary output image.
showasImageNovi.mNormalizes the image to a range [0,255].
SPS_novi.mImplements the Liu et al. approach.
undistort_WAHRSIS_imgs.pyUndistorts our sky camera images; needed during the creation of the dataset.
The various functions required in SLIC superpixel segmentation can be found in the folder
./SegmentationToolbox. The core functions of SLIC are re-distributed under GNU General Public License terms.
In addition to all the related codes, we have also shared the generated results. These files are contained in the folder
Please run the following to generate the various figures and tables in the paper.
Figure1.mDemonstration of the proposed segmentation algorithm.
Figure3.mComputes the cloud coverage of the sample images of the dataset.
Figure6.mPerformance of the various color channels for nighttime image segmentation.
Statistics of SWINSEG dataset.ipynbComputes the distribution of images in the image dataset.
Table2.mPerformance evaluation of various benchmarking algorithms.