Simple image transformations—resizing and flipping

Now we're able to load an image, it's time to do some simple image processing. The operations we're going to review—resize and flip—are basic and usually used as preliminary steps of complex computer vision algorithms.

Getting ready

You need to have OpenCV 3.x installed with Python API support.

How to do it...

For this recipe, we need the following steps to be executed:

Load an image and print its original size:

img = cv2.imread('../data/Lena.png')
print('original image shape:', img.shape)

OpenCV offers several ways of using the cv2.resize function. We can set the target size (width, height) in pixels as the second parameter:

width, height = 128, 256
resized_img = cv2.resize(img, (width, height))
print('resized to 128x256 image shape:', resized_img.shape)

Resize by setting multipliers of the image's original width and height:

w_mult, h_mult = 0.25, 0.5
resized_img = cv2.resize(img, (0, 0), resized_img, w_mult, h_mult)
print('image shape:', resized_img.shape)

Resize using nearest-neighbor interpolation instead of the default one:

w_mult, h_mult = 2, 4
resized_img = cv2.resize(img, (0, 0), resized_img, w_mult, h_mult, cv2.INTER_NEAREST)
print('half sized image shape:', resized_img.shape)

Reflect the image along its horizontal x-axis. To do this, we should pass 0 as the last argument of the cv2.flip function:

img_flip_along_x = cv2.flip(img, 0)

Of course, it's possible to flip the image along its vertical y-axis—just pass any value greater than 0:

img_flip_along_y = cv2.flip(img, 1)

We can flip both x and y simultaneously by passing any negative value to the function:

img_flipped_xy = cv2.flip(img, -1)

How it works...

We can play with interpolation mode in cv2.resize—it defines how values between pixels are computed. There are quite a few types of interpolation, each with a different outcome. This argument can be passed as the last one and doesn't influence the result's size—only the quality and smoothness of the output.

By default, bilinear interpolation (cv2.INTER_LINEAR) is used. But in some situations, it may be necessary to apply other, more complicated options.

The cv2.flip function is used for mirroring images. It doesn't change the size of an image, but rather swaps the pixels.