Image Processing with R |

Image processing in R, using packages like imager, enables powerful analysis and manipulation of digital images. This blog demonstrates key techniques—enhancement, denoising, and histogram equalization—using the imager package, with visual results to illustrate their effects, based on concepts from the “Image Data Analysis” document.

Understanding Digital Images

A digital image is a 2D function $f(x, y): \mathbb{R}^2 \rightarrow \mathbb{R}$, where $(x, y)$ are spatial coordinates, and $f(x, y)$ represents pixel intensity. In digital images, pixels have discrete locations and values. For 8-bit grayscale images, intensities range from 0 (black) to 255 (white). Images are categorized as:

Binary: Pixels are 0 (black) or 1 (white).
Grayscale: Pixel values in ${0, \ldots, 255}$.
Color: Three channels (RGB), each in ${0, \ldots, 255}$.

Image Processing with `imager`

The imager package, built on CImg, supports loading, manipulating, and visualizing images. Below, we demonstrate key operations with code and visual outputs using the built-in “parrots.png” image.

1. Image Acquisition

Load and display an image:

library(imager)
file <- system.file('extdata/parrots.png', package='imager')
img <- load.image(file)
plot(img, main="Original Parrots Image")

Result: Original Parrots Image

2. Image Enhancement (Blurring)

Blurring enhances images for specific applications, e.g., smoothing details:

img_blurry <- isoblur(img, sigma=10)
plot(img_blurry, main="Blurred Image (sigma=10)")

Result: Blurred Parrots Image Note: The blurred image appears smoother, reducing fine details like feather textures.

3. Image Denoising

Denoising removes noise while preserving structure. Add noise and apply anisotropic blurring:

img_noisy <- img + 0.5 * rnorm(prod(dim(img)))
img_denoised <- blur_anisotropic(img_noisy, ampl=1e3, sharp=0.3)
layout(t(1:2))
plot(img_noisy, main="Noisy Image")
plot(img_denoised, main="Denoised Image (Anisotropic)")

Result: Noisy vs Denoised Note: The noisy image shows random speckles, while the denoised image restores clarity, preserving edges.

4. Histogram Equalization

Histogram equalization enhances contrast by redistributing pixel intensities:

img_gray <- grayscale(img)
f <- ecdf(img_gray)
img_equalized <- f(img_gray) %>% as.cimg(dim=dim(img_gray))
layout(t(1:2))
plot(img_gray, main="Grayscale Image")
plot(img_equalized, main="Histogram Equalized Image")

Result: Grayscale vs
Equalized Note: The equalized image has improved contrast, making details like color variations more distinct.

5. Morphological Processing

Thresholding segments objects by intensity:

img_gray <- grayscale(img)
threshold(img_gray, "20%") %>% plot(main="Thresholded Image (20%)")

Result: Thresholded Image Note: Thresholding creates a binary image, highlighting brighter regions (e.g., white feathers) against darker ones.

6. Putting Everything Together

library(imager)

# Set up directory for saving plots (optional: create if it doesn't exist)
output_dir <- "plots"
if (!dir.exists(output_dir)) {
  dir.create(output_dir)
}

# Image Acquisition
file <- system.file('extdata/parrots.png', package='imager')
img <- load.image(file)
plot(img, main="Original Parrots Image")
dev.copy(png, file.path(output_dir, "original_parrots.png"))
dev.off()

# Image Enhancement (Blurring)
img_blurry <- isoblur(img, sigma=10)
plot(img_blurry, main="Blurred Image (sigma=10)")
dev.copy(png, file.path(output_dir, "blurred_parrots.png"))
dev.off()

# Image Denoising
img_noisy <- img + 0.5 * rnorm(prod(dim(img)))
img_denoised <- blur_anisotropic(img_noisy, ampl=1e3, sharp=0.3)
layout(t(1:2))
plot(img_noisy, main="Noisy Image")
plot(img_denoised, main="Denoised Image (Anisotropic)")
dev.copy(png, file.path(output_dir, "noisy_vs_denoised_parrots.png"))
dev.off()

# Histogram Equalization
img_gray <- grayscale(img)
f <- ecdf(img_gray)
img_equalized <- f(img_gray) %>% as.cimg(dim=dim(img_gray))
layout(t(1:2))
plot(img_gray, main="Grayscale Image")
plot(img_equalized, main="Histogram Equalized Image")
dev.copy(png, file.path(output_dir, "grayscale_vs_equalized_parrots.png"))
dev.off()

# Morphological Processing
img_gray <- grayscale(img)
threshold(img_gray, "20%") %>% plot(main="Thresholded Image (20%)")
dev.copy(png, file.path(output_dir, "thresholded_parrots.png"))
dev.off()

Applications

Image processing in R is used in: - Automotive: Lane detection, obstacle warning. - Medical: Diagnostic imaging, surgical assistance. - Security: Face recognition, surveillance. - Media: Special effects, image editing.

Conclusion

The imager package in R simplifies image processing tasks like enhancement, denoising, and histogram equalization. Visual results demonstrate how these techniques transform images, improving quality or extracting features. Explore imager and imagerExtra for advanced applications.

Resources: