Lecture 4 Image Processing in the spatial domain

Lecture 4 Image Processing in the spatial domain

Spatial domain enhancement • Enhance features in the input image for further processing – What features? – Application dependant. • Collection of enhancing algorithms that work directly on pixels are called Spatial domain enhancement.

Zooming an image • Given an image A of size m 1 X n 1, resample A to get image of size m 2 X n 2. • How to compute grey-level values at new points? 4 x 4 grid 10 x 10 grid

Resampling by integer factor • m 1 x n 1 = 4 x 4, m 2 x n 2 = 10 x 10 • Pixel replication

Pixel Replication

Non-integer factor resampling

Nearest neighbor interpolation • Pixel replication is a special case of NN.

Nearest neighbor interpolation Original 200 x 200 Resampled from 128 x 128 Resampled from 64 x 64 Resampled from 32 x 32

Bilinear interpolation y 1 x y

Bilinear Interpolation Original 200 x 200 Resampled from 128 x 128 Resampled from 64 x 64 Resampled from 32 x 32

Shrinking • Shrinking can be done effectively in the same way. • Might need some low pass filtering to avoid aliasing.

Grey-level Transformations • T depends only on the grey-level value of a pixel • Since it does not depend on neighborhood pixels, this is called point processing

Thresholding • s 255 0 a 255 r

Image negative • s 255 0 255 r

Grey-level value transformation

Log transform