Some of these will be transfered periodically to Open issues.
- Binary packages
- Performance tests for OTB (orthorectification for different sensors, reading and writing polygons,...) to be able to assess impact of modifications (OTB code and external librairies).
- Code review for GDALImageIO
- Code review for DrawPathList and integration of DrawPointSetImageFilter (general reflexion is required on displaying vector data)
- Review JPEG2000 usage to get correct performances
- Wavelet transform (streaming !)
JPG and PNG reading could be done with gdal to handle associated georeference information
ImageReader should be capable to access the different images when the product contains several images (cf bug #20). Note: there is several possibilities: return an image list, add a parameter to the ImageReader to specify which image to return, etc.
Remove libSVM output during learning (.........*......)
- Continuous for a Windows platform
- Investigate FFTW problem when running multiple instances of the otb::OverlapSaveConvolutionImageFilter.
- Integrate the Geometry Library.
Test the Pleaides format (JP2K & DIMAP) available in LargeInput/PLEIADES --> 2 products are available (SENSOR and ORTHO)
- Test the reading of the grass format (cf mail on otb-user)
- Consistent and homogeous channel numbers in OTB methods - The first channel is number 0 or number 1?
Move the SpatialObjectDXFReader closer to the VectorDataIO hierarchy.
- Add light Smoke test suite: about 100 tests, <1min compilation, <1min execution, at least one per otblib, without the need for otb-data (possibility to use otb logo in .h and checksum on the output image). ON by default.
Extend DEM management (geotiff, IGN DEM format...). Automatic binary packages generation of Linux/windows/Mac (stable and development version)
- ImageListToVectorImage (and vectorimagetoimagelist) must be threaded
- Create the '--compare-image-estimators' in order to compute and compare estimators (MSRN,RMSE,...) between baseline and result.
- Missing tests
- Remove the INCLUDE_DIRECTORIES from the main cmakelists and move them with the corresponding target_link_librairies (ex: include for mapnik in Code/Common/CMakeLists.txt) to avoid full recompilation.
Update to libsvm 3.0
- Add contribution on GPU programming
Integrate dev version of libLAS (which compiles fine with gcc4.5 and now has a CMake based build system) : fix issues with libLAS developers about GDAL/GeoTiff import
- Code review of otbImageFileReader : avoid multiple calls of ossimImageHandler and getImageGeometry ( time consuming for Spot5 model)
Profiling upon the computation of covariance shows that StreamingStatisticsVectorImageFilter is not effective (ffedback from L Mercier LPG Grenoble)
- OTB must be able to work with the maximum of external libraries : ITK4.x, OSSIM...
- OTB internal applications must handle extractROI on their output. (give the possibility to add tests about Optical and SAR calibration applications)
- Examples to add in the SG
- Right Angles
- Indices Soil et Water
- Finish lines --> bright/dark selection
- Epipolar geometry
- Formosat sensor model
- Pléiades sensor model
- Geoeye sensor model
Redesign and enrich the commandLineArgumentParser in a more generic way to be able to produce other type of GUI (Monteverdi modules, Zoo Project config file : link zoo_project).
- Finish User guide Documentation ( description for each module)
- Zoom in the Fine correlation module
- Export/Import workflow in Monteverdi
Convert Monteverdi to a QGIS plugin
- Add local descriptors on regions in the Feature extraction module (Flusser, skewness, kurtosis, mean,elongation, variance, principal axis...)
- QT GUI
Color mapping module (using OTB color maps filter) Module to import a DEM image in a given projection
- Smoothing module with different methods (done in Applications)
- Road extraction module
- Warping with respect to a deformation field
- Spectral Viewer : add input text selection to band selection sliders
- Monteverdi pipeline could carry static objects: standard quick look, scenario...
- Add a close dataset button in monteverdi
- wrap StreamingStatisticsVectorImageFilter (last try : error when wrapping vnl_svd due to inline function)
- Draw shapefiles with country borders in the image extent tab of the ortho-rectification application: Can use a vector layer over the viewer components
- Use Otsu thresholding in the object counting application
- Import examples into the object counting application
- Network extraction application: Display segments with a different color according to the input type used
- Network extraction application: Fix the filter updating problem (for now pointer is reinitialized for each computation)
- Supervised classification application : add active learning, model selection and validation, classifier fusion.
- Easy translation between otbVectorData and PolygonList, pointSet, pathList etc.
- Simple concat of images and more
- Soil line estimation
- Add streaming capabilities to DEMToImageGenerator and DEMToOrthoImageGenerator. This should be an easy case (no boundaries) ideal for somebody who wants to get familiar with the streaming process in OTB.
- Add the possibility to display a RGB composition from the features list in otbFeatureExtractionAppli.
- Create a DEM characteristics generator (slope, aspect, etc.)
- Integrate random convolution-based disparity map estimation in library and in fine registration application
Read a VectorImage of type std::complex<>
Support for hdf format
Viewer : Handle the drag on the Full widget
- Trees and Tree Stand Extraction Application
SAR polarimetry analysis tools
- NL-means filtering. Code is available here, but it may be easier to write a filter from this commented pseudocode. Also note that a hashed version exists (see "Hashed Non-Local Means for Rapid Image Filtering", Dowson, N ; Salvado, O ;)
- dct denoising : http://www.ipol.im/pub/algo/ys_dct_denoising/
- Add concept of networks in OTB (roads, hydro...)
- Automatic download of Aeronet files (read image metadata place and time)
- Automatic download of DEM tiles
- Classification without resampling of data with different resolutions
- Automatic DEM generation: orthorectification -> correlation -> filtering -> TIN generation
- Manage variogram (geostatistics)
- Cut images from a buffer
- Apply segmentation and classification on masked image (venus contribution?) -> Study direct solutions via gdal http://trac.osgeo.org/gdal/wiki/rfc15_nodatabitmask
- Perform multi feature Mean Shift with different parameters for each layer
- Add GetQuickLook in ImageFileReader to manage multi resolution pyramids (OSSIM, GDAL, overview format)
- Boosting in OTB
- Particle filter in OTB
- Fusion of classifier Framework
- Fast LevelSet Transform (contribution from E.Dubois)
- Data/image whitening filter in OTB (related to ICA)
- Active learning based on margin clustering
- True Texture Index: texture index based on speckle analysis for SAR images
- MultiScale Isotropic Attention detector
Maximum Autocorrelation Factor decomposition (and kernelized version) (Matlab/Envi code available here) Multivariate Alteration Detection (Matlab/Envi code available here)
- Multi-resolution spatial unmixing framework
- Sparse Matrix Transform
- Random Forest Classifier
- Galib based genetic algorithm optimizer
Add class to compute automatically orthorectification parameters (integrate in Applications and Monteverdi)
- Find a way to handle big images in OBIA framework (Automatic caching of input segmented image? Look at contribution from J.Berthrand)
- OBIA : export full scene result in PostGIS
- Investigate the potential use of Waffles C++ API or OpenCV as a backend for machine learning tasks
- OffsetEstimation (M. Seymour)
- SensorGroundPosition (M. Seymour)
- Line detection (bright/dark lines and parallel lines) (Jan Wegner)
Fisher classification (Jan Wegner)
- Raster attribute table support (Alexander Herzig)