This will be a short presentation how a procedural image views can be used. We will generate an image which consists of multiple checkerboards with different tile size being super-positioned over each other and summed.
The BoltView library provides an purely procedural image view which behaves like checker board image view, although the checker board image does not exist in memory. The view computes proper value on demand when executing the
operator. The library also provides overloads of arithmetic operators, which can be used for basic manipulation with image views, although the view they return also does not have data…
In this article we will explore the basic usage of the BoltView concepts to generate image of the famous Mandelbrot’s set fractal. We will start with simple version running on CPU — version useful for algorithm debugging. After that we will modify the code to run on GPU.
The first step is to prepare a place to store the actual image data. For this purpose we will create an instance of
bolt::HostImage template. It is a class following the RAII (Resource Aquisition Is Initialization) paradigm. The two template parameters are the type of pixel values (
Vector<uint8_t, 3> -- to…
We are happy to announce that we are doing the first public release of a library called BoltView (web, github). It was developed as an internal tool in Eyen SE for image processing and easy parallelization on CUDA-enabled graphic cards.
The library was so far successfully used in various projects mostly oriented on processing volumetric (voxel) data — filtering, segmentation, tomography reconstructions, visualization, etc.
This blog is dedicated to the C++ library called BoltView, which provides data-structures and algorithms for image processing in CUDA.