Figure Processing

Image Compression

I have tried an excellent PNG file compressor online: TinyPNG. Usually this can compress a figure by about 60% without lossing visually detectable quality! This should always be used whenever possible.

Unlike JPEG, PNG doesn’t typically have a lossy compression scheme. What TinyPNG can achieve is something I haven’t really found an alternative for. I made a donation of $5 to them. Great work!

• ffmpeg has some built-in compressor when combining figures into movies, so it make less sense to compress each figure before passing them to ffmpeg.
• When merging images, the size reduction is lost for the combined image because 24-bit colors is the default settings for PNG savings. We need to optimize the image again.

This video explains many things about the differences between PNG and JPEG:

So basically, if you can count the number of colors in your image (e.g. most scientific plottings), use PNG. Even though PNG supports 24-bit colors, 8-bit (256 colors) is enough for most scientific images. This is the most important trick used in TinyPNG.

Image Manipulation

There are many online image merging and splitting tools available. However, we can also process images by writing some codes in Julia! This can be easily achieved with Images.jl.

Basics

In Julia, images as multidimensional arrays are stored in column-major order, which means that for an image with size size(img) = (512, 768), the first index corresponds to the vertical axis (column, height) and the second to the horizontal axis (row, width).

To use Images.jl, add using Images to the beginning of the script.

Cropping

The syntax is

img_source[y1:y2, x1:x2]

For example, to crop an image from sides by 1/8 of img_source each side and leave it as it is from top to bottom,

img_cropped = @view img_source[ :,floor(Int, 1/8*img_size[2]) : floor(Int, 7/8*img_size[2])]

Resizing

Resizing is a method to resize an image to a given specific output image shape.

img_square = imresize(img_source, (400, 400))

Rescaling

Rescale operation resizes an image by a given scaling factor. The scaling factor can either be a single floating point value, or multiple values - one along each axis. Image scaling is the process of changing the size of an image while preserving the original aspect ratio.

img_small = imresize(img_source, ratio=1/4);
img_short = imresize(img_source, ratio=(1/4, 1));

Merging

• Layout: 1 on the left, 2 rows on the right

dir = "./"
ima = load(dir*"1a.png")
imb = load(dir*"1b.png")
imc = load(dir*"1c.png")
sa = size(ima)
im_right = vcat(imb, imc) # 1b and 1c have the same sizes
sright = size(im_right)

# Resize the image
percentage_scale = sa[1] / sright[1]
new_size = trunc.(Int, sright .* percentage_scale)
im_right_rescaled = imresize(im_right, new_size)

im = hcat(ima, im_right_rescaled)
save("1.png", im)

• Layout: 2 rows * 2 columns

dir = "./"
ima = load(dir*"2a.png")
imb = load(dir*"2b.png")
imc = load(dir*"2c.png")
imd = load(dir*"2d.png")
sa = size(ima)
sc = size(imc)
# Resize the bottom image
percentage_scale = sa[2] / sc[2]
new_size = trunc.(Int, sc .* percentage_scale)
im_rescaled = imresize(imc, new_size)

im_left = vcat(ima, im_rescaled) # 1a and 1b have the same sizes
im_rescaled = imresize(imd, new_size)

im_right = vcat(imb, im_rescaled) # 1a and 1b have the same sizes

im = hcat(im_left, im_right)
save("1.png", im)