FPGA 2D filter

This FPGA-based image filter can filter a provided image through several methods.

Each of these filters have been tested on the standard test image Lenna.

Implemented filters

Averaging

Each pixel (except the top and bottom ones) is replaced by the average of all its surrounding pixels.

Vertical contour detection filter

This filter applies factors from the following matrix, in order to detect vertical countours:

__         __
| -1   0  +1 |
| -2   0  +2 |
| -1   0  +1 |
‾‾         ‾‾

Horizontal contour detection filter

This filter applies factors from the following matrix, in order to detect horizontal countours:

__         __
| -1  -2  -1 |
|  0   0   0 |
| +1  +2  +1 |
‾‾         ‾‾

Full contour detection filter

This filter applies factors from the following matrix, in order to detect all countours:

__         __
|  0  -1   0 |
| -1   4  -1 |
|  0  -1   0 |
‾‾         ‾‾

Implementing new filters

This software has been designed with modularity in mind, up to a certain point.

That means you can straightforwardly add new filter matrixes to filter.vhd, or replace some existing ones.

Testing

To test the filters, you can use the test bench (tb_filter.vhd) as a Vivado simulation source.

In that file, depending on the filtering method you want to test, you should uncomment the relevant value for filter_type_to_test and comment out the others.

In fpga-2d-filter.sim/sim_1/behav/xsim, you should provide an input.dat file containing each pixel, line by line, in binary format.

The test bench will output the filtered image, in that same directory, into output.dat.

You may convert this input data to a bitmap image, manually, by running dat2bmp.m with GNU Octave. This will create output.bmp at the root of the project!

___________________

This software has been designed by Firmin Launay, in 2024, as part of the Reconfigurable Architectures course at Polytech Dijon.