## Image Convolution

## Learning Objectives * Learn about image convolutions and what makes them a good problem for solving on a GPU * Learn what a naive image convolution may look like

#### Image convolution

Over the next few lectures we will be looking at some common GPU optimizations with an image convolution as the motivational example.

* A good problem to solve on a GPU. * Can take advantage of a number of common optimizations. * Convolution is a very powerful algorithm with many applications. * Deep neural networks. * Image processing.

#### Why are image convolutions good on a GPU?

* The algorithm is **embarrassingly parallel**. * Each work-item in the computation can be calculated entirely independently. * The algorithm is computation heavy. * A large number of operations are performed for each work-item in the computation, particularly when using large filters. * The algorithm requires a large bandwidth. * A lot of data must be passed through the GPU to process an image, particularly if the image is very high resolution.

#### Image convolution definition

![SYCL](../common-revealjs/images/image_convolution_definition.png "SYCL")

#### Image convolution definition

![SYCL](../common-revealjs/images/image_convolution_definition.png "SYCL")

* A filter of a given size is applied as a stencil to the position of each pixel in the input image. * Each pixel covered by the filter is then multiples with the corresponding element in the filter. * The result of these multiplications is then summed to give the resulting output pixel. * Here we have a 3x3 gaussian blur approximation as an example.

#### Image convolution example

![SYCL](../common-revealjs/images/image_convolution_example.png "SYCL")

#### Image convolution data flow

![SYCL](../common-revealjs/images/image_convolution_data_flow.png "SYCL")

* We have a single kernel function. * It must read from the input image data and writes to the output image data. * It must also read from the filter. * The input image data and the filter don't need to be copied back to the host. * The output image data can be uninitialized.

#### Implementation

* We provide a naive implementation of a SYCL application which implements the image convolution algorithm. * This will be the basis for optimization in later lectures and exercises. * The implementation uses the stb image library to allow us to visualize our results. * The implementation also uses a benchmark function to allow us to measure the performance as we make optimizations.

#### Reference image

![SYCL](../common-revealjs/images/dogs.png "SYCL")

* We provide a reference image to use in the exercise. * This is in Code_Exercises/Images * This image is a 512x512 RGBA png. * Feel free to use your own image but we recommend keeping to this format.

#### Input/output image locations

auto inputImageFile = "../Code_Exercises/Images/dogs.png";
auto outputImageFile = ../Code_Exercises/Images/blurred_dogs.png";

* The reference code and the solutions to the remaining exercises use these strings to refernce the location of the input and output image. * Before compiling these you will have to update this to point to the image in the development environment.

#### Convolution filters

auto filter = util::generate_filter(util::filter_type filterType, int width);

* The utility for generating the filter data takes a `filter_type` enum which can be either `identity` or `blur` and a width. * Feel free to experiment with different variations. * Note that the filter width should always be an odd value.

#### Naive image convolution performance

![SYCL](../common-revealjs/images/image_convolution_performance.png "SYCL")

## Questions

#### Exercise

Code_Exercises/Image_Convolution/reference

Take a look at the naive image convolution code provided and familiarize yourself with it.