In neuroconductor-devel/Rvision: Basic Computer Vision Library

library(Rvision)

This page describes all basic operations on videos and images.

---

• 3.1 - Object properties
• 3.2 - Displaying images
  • 3.2.1 - Method compatible with base R
  • 3.2.2 - Method incompatible with base R
• 3.3 - Accessing/replacing parts of an image

---

3.1 - Object properties

| | Video | Stream | Image | VideoWriter | |:--- |:------- |:-------- |:------- |:------------- | | Testing for | isVideo(x) | isStream(x) | isImage(x) | isVideoWriter(x) | | Dimensions | dim(x) | dim(x) | dim(x) | - | | Number of rows | nrow(x) | nrow(x) | nrow(x) | - | | Number of columns | ncol(x) | ncol(x) | ncol(x) | - | | Number of channels | - | - | nchan(x) | - | | Number of frames | nframes(x) | - | - | - | | Framerate | fps(x) | - | - | - | | Codec | codec(x) | - | - | - | | Bit depth | - | - | bitdepth(x) | - | | Color space | - | - | colorspace(x) | - |

---

3.2 - Displaying images

There are two methods for displaying images. The first method offers full compatibility with base R plotting methods, which allows for drawing non-destructively over an image using graphics::lines(), garaphics::points(), etc. This method is however is however rather slow and therefore not suitable for applications requiring quick refresh rates between successive images (e.g. when playing a video).

The second method uses OpenCV's capabilities to display images at high frame rates, making it suitable to play videos for instance. It is however incompatible with base R plotting functions. However you can use the drawing functions provided with Rvision to draw over images - destructively this time - before displaying them.

3.2.1 - Method compatible with base R

path_to_image <- system.file("sample_img", "bunny.png", package = "Rvision")
my_image <- image(filename = path_to_image)

plot(my_image)
abline(h = 400, col = "red", lwd = 5)
abline(h = 550, col = "red", lwd = 5)
abline(v = 430, col = "red", lwd = 5)
abline(v = 570, col = "red", lwd = 5)

3.2.2 - Method incompatible with base R

path_to_video <- system.file("sample_vid", "Balloon.mp4", package = "Rvision")
my_video <- video(filename = path_to_video)

# Create an empty display window (note: it can hide behind other windows)
newDisplay("My display", nrow(my_video) / 2, ncol(my_video) / 2) 

# Display images (note: it creates the display window if it doesn't exist yet)
# The 3rd argument ('delay') is the minimum time in ms during which the image 
# should be displayed before it can be replaced
for (i in 1:25) {
  display(readNext(my_video), "My display", 25, nrow(my_video) / 2, ncol(my_video) / 2)
}

# Close display
destroyDisplay("My display")

# Close all opened displays
destroyAllDisplays()

---

3.3 - Accessing/replacing parts of an image

You can access parts of an image in the same way you would do it with matrices and arrays in base R.

my_image[]          # Entire image
my_image[1, 1]      # Bottom-right pixel
my_image[1, ]       # Bottom row of pixels
my_image[, 1]       # Leftmost column of pixels
my_image[1:5, 1:5]  # All pixels between the 1st and 5th row and column
my_image[c(TRUE, FALSE), c(TRUE, FALSE)]  # Every other row and column of pixels

Note that images are not exactly like traditional matrices. Color images have multiple layers, each corresponsing to a dimension in the BGR or BGRA space (Blue, Green, Red, Alpha transparency). Therefore, when accessing a pixel of a color image, you will obtain 3 or 4 values in return. Grayscale images, on the other hand, function similarly to regular monolayer matrices in base R.

To replace parts of an image, you can use the <- assignment operator. With grayscale images, it works exactly like in base R with matrices.

my_gray_image <- changeColorSpace(my_image, "GRAY")
my_gray_image[1:250, 1:250] <- 0  # Turn the corresponding pixels to black
plot(my_gray_image)

With color images however, you will need to supply 3 (for BGR images) or 4 (for BGRA images) values (or any multiple of 3 or 4 respectively).

my_image[1:250, 1:250] <- c(0, 0, 255, 255) # Turn the corresponding pixels to red
plot(my_image)

The utility function col2bgr provided by Rvision (not to be confused with col2rgb provided by the grDevices package and on which col2bgr is based) will conveniently convert any of the R-compatible color specifications to color codes compatible with Rvision images.

my_image[1:250, 1:250] <- col2bgr("blue", alpha = TRUE) # Turn the corresponding pixels to blue
plot(my_image)

neuroconductor-devel/Rvision documentation built on May 16, 2021, 5:16 p.m.