astex-icube / astex Goto Github PK

License: GNU Lesser General Public License v2.1

CMake 2.19% C++ 93.28% C 4.47% Shell 0.01% Makefile 0.01% GLSL 0.06%

astex's Introduction

ASTex

ASTex is an open-source library for texture analysis and synthesis. The term “texture” must be understood as in computer graphics. The purpose of this library is to support collaborative coding and to allow for comparison with recently published algorithms.

Installation

Dependencies:

To compile and use ASTex, you need some libraries:

ITK 5. min
zlib
openexr for saving/loading images in floating point format.

You need some classic development tools (minimal supported version)

git
cmake 3.0 min
a recent C++ compiler
- g++ 6
- clang 3.3
- Visual Studio C++ 2017 /2019

Linux

Just install packages:

libpng-dev
libopenexr-dev And get github version of InsightToolkit, compile & install

Mac OS/X

The most simple way to install dependencies is to use homebrew package system. Then you can install the dependencies:

brew install insighttoolkit
brew install openexr
brew install libpng

Windows

Softwares:

VisualStudio C++ (2017 min)
Git
CMake (3.0 min)

install VCPKG For ASTex

open a power-shell
git clone https://github.com/Microsoft/vcpkg.git
var env VCPKG_DEFAULT_TRIPLET=x64-windows
cd vcpkg
.\vcpkg\bootstrap-vcpkg.bat
.\vcpkg.exe install itk openexr
Remarque: ATTENTION LA COMPILATION NECESSITE X.X Go D'ESPACE DISQUE ! Vous pouvez enlever les répertoires buildtrees, downloads et packages

Compile ASTex

launch CMake, chose src dir and binary dir
Specify tool chain file: XXXX/vcpkg_ast/scripts/buildsystems/vcpkg.cmake
Configure again
Generate
Launch Solution
Enjoy

Data

Some tests, tutorials and algorithms use read example images and write some results. In order to keep original Data directory of ASTex clean we use a copy which pass can be choosen at cmake configuration stage.

You have to copy yourself the contain of the Data directory in to the right place (see ASTEX_TEMPO_PATH clake variable).

Compilation

on Linux & Mac

Use CMake as usual:

create a build directory as same level than ASTex (ASTex-build or ASTex-buildDebug for example)
go inside build directory and do cmake ../ASTex (or use gui)
or let (a recent) Qtcreator do the job !

CMake Options

There are some original options/values that can be set at the cmake stage:

ASTEX_ALGO_xxx choose to build the different implemented algorithms.
ASTEX_BUILD_xxx choose to build bench/tuto/test
ASTEX_PERSO_xxx for each directory added in ASTex that contain a CMakeLists.txt set this to ON to build. When you add a directory just relaunch cmake.
ASTEX_TEMPO_PATH path of directory use to store images for test and tuto (copy ASTex/Data into it)

Contributing to ASTex

All contributions to ASTex are welcome and will be examined.

To contribute:

you need a github account.
fork ASTex repository on your account
create a branch from the develop one
develop your contribution
do pull-request on ASTex/develop branch

astex's People

Contributors

Stargazers

Watchers

Forkers

sylvainthery frederic-larue drlutzi kvanhoey pandinosaurus jmdischler basilesauvage nathanielseyler kumaa catherineytw creatorlars peterzhousz romimap chgrenier jackzhousz peterzs qwendling

astex's Issues

itkFastSimpleMutex with ITK

Hello,

It seems the latest ITK does not support itkSimpleFastMutexLock.h. Is there a quick solution?

csv

Image type with variable number of channels

We should create an ImageType that allows the user to have a variable number of channels, for example one comprised of red, green, blue, alpha, normal, and height. This would be useful to extend the scope of ASTex to various textures. One would be able to apply classic operators such as * or +.

Create a custom ImageType with a variable number of channels

Mathematical operations on images

It would be great to have some of the classical operations on image processing such as adding two images of the same size, point-wise multiplication and matrix multiplication.
They are relatively easy and can also be implemented in place of Itk filter (such as itk::NaryAddImageFilter) if necessary.

The "save" function of real Image types produces empty files when called with the default syntax

Real-valued image types, such as ImageGrayd and ImageRGBd, produce empty files when called with the default syntax (save(filename)). This enforces the use of IO::save01_in_u8(image, filename), but this is inconvenient with template types that can be different Image types.

I suggest to change save(filename) so that it saves real-valued images by assuming their value is between 0 and 1, and clamp values that are not in this range.

Change the behavior of the save function to make it usable by templates

ImageType::initItk(w, h, true) does not always initialize a null-valued image

Everything is in the title. I have stumbled upon this issue only for ImageRGB types, though.

ConstImageTypeX cannot be deduced from ImageTypeX

The following is only a minor inconvenience, and I have provided a workaround on the bottom of the issue.

Probably because of how itk works, the types for images are not the same between their non-const and const version, so instead of using const ImageX, we use ConstImageX. The current problem with this syntax is that the const version of an image cannot be deduced, which can enforce the use of several redundant template members.

Example: in the following code, I want to make a itk filter that inputs an image using an ASTex Image type as a template. The following is perfectly fine:

template<typename TInput> class FilterGetisGI : public itk::ImageToImageFilter<typename TInput::ItkImg, typename TInput::ItkImg>

But it enforces that TInput is a non ConstImageType, because the output of the filter is necessarily non ConstImageType as well. I would like to use something like:

template<typename TNonConstImage> class FilterGetisGI : public itk::ImageToImageFilter<typename TConstImage::ItkImg (?), typename TNonConstImage::ItkImg>

but it is not currently possible to deduce the Const version of an image from its NonConst version. A current workaround is to use something like:

template<typename TNonConstImage, TConstImage> class FilterGetisGI : public itk::ImageToImageFilter<typename TConstImage::ItkImg, typename TNonConstImage::ItkImg>

which forces the user to provide the Const version of the image type with the NonConst.

Add a way to deduce const image types from non const image types
Add a way to deduce non const image types from const image types