In this tutorial we use the ColorFiltering module of JeVois to interactively explore how different image filters may affect an image, using a live video stream and live processing inside the JeVois smart camera.

This tutorial is intended as a fun, interactive introduction to image processing for K-12 / STEM (kindergarten to high-school, science, technology, engineering and maths) students or anyone interested in learning more about computer vision and image or video processing.

JeVois v1.3.2

Setting up

Get the latest microSD disk image from http://jevois.org/start and flash it.
Connect JeVois to a host computer and allow it to start up.
Open a video viewer on your host computer and select YUYV 640x240 @ 30 fps (see JeVois-A33 Quickstart user guide)
Open a serial communication to JeVois (see Command-line interface user guide) or use jevois-cmd (see Connecting to JeVois using serial-over-USB: Linux host) to send the commands below to JeVois.

Operation

The ColorFiltering module implements an number of image filters. The original video captured by JeVois is shown to the left of the display, and the results of applying the selected filter is shown to the right. The parameter settings of the currently running filter are shown as an overlay on top of the filter results at right.

The module has one main parameter that determines which filter to use: effect with values: NoEffect, Blur, Median, Morpho, Laplacian, and Bilateral.

When a given effect is selected, the corresponding image filter is implemented. Different image filters may have additional parameters. For example, the Median filter has a parameter ksize to set the kernel size; the Morpho filter has a parameter op to select the morphological operation (erosion, dilation, opening, closing, top hat, etc).

Thus, the typical way to use this module is:

set an effect with setpar effect Blur, setpar effect Median, etc
type help to see what additional parameters have become available that are specific to the currently selected effect (those ara slso displayed as an overlay on top of the processed image at right)
further set these parameters using additional setpar commands.

In what follows we explore the different filters and their settings.

Effect: Blur

This filter replaces each pixel's value by the average of all pixel values in a rectangular box around the pixel of interest. This leads to a blur effect on the image, more pronounced with bigger box sizes.

See http://docs.opencv.org/3.2.0/d4/d86/group__imgproc__filter.html#ga8c45db9afe636703801b0b2e440fce37

setpar effect Blur

Blur filter with default settings

setpar ksize 11 3

Blur filter with kernel size 11x3

setpar ksize 3 11

Blur filter with kernel size 3x11

setpar ksize 11 11

Blur filter with kernel size 11x11

Effect: Median

This filter replaces each pixel by the median value within a square box around that pixel. Tends to blur and remove salt-and-peper noise from the image.

See http://docs.opencv.org/3.2.0/d4/d86/group__imgproc__filter.html#ga564869aa33e58769b4469101aac458f9

setpar effect Median

Median filter with default settings

setpar ksize 7

Median filter with kernel size 7x7 smoothes out the image

Median filter with kernel size 7x7 smoothes out a texture

setpar ksize 13

Median filter with kernel size 13x13 smoothes out the image

Effect: Morpho

This filter applies mathematical morphology operations, which aer operations in which every pixel value is replaced by the result of some operation applied within a shaped neighborhood (the morphological structuring element) shifted to the location of the pixel of interest.

See http://docs.opencv.org/3.2.0/d9/d61/tutorial_py_morphological_ops.html for an introduction, and http://docs.opencv.org/3.2.0/d4/d86/group__imgproc__filter.html