Computer image processing with example

Computer image processing means applying algorithms to digital images to improve them, extract information from them or prepare them for storage and transmission. An image is a 2-D array of pixel values, and processing converts one array into another (or into data).

Main areas (Steinmetz)

• Image synthesis: creating images from descriptions or models.
• Image analysis: extracting features and recognising objects.
• Image transmission: sending images efficiently over a network.

Typical processing steps

 Acquire -> Pre-process -> Enhance -> Segment -> Represent
 (scan)     (noise,         (contrast,  (find      & output
             geometry)       sharpen)    objects)

1. Acquisition: capture or scan the picture and digitise it.
2. Pre-processing / restoration: remove noise and blur, correct geometric distortion.
3. Enhancement: improve visibility by contrast stretching, histogram equalisation, sharpening or smoothing filters.
4. Segmentation: separate objects from background by thresholding or edge detection.
5. Representation and description: express objects with boundaries or features.
6. Compression and output: store, print or display.

Operations by type

• Point operations: each pixel changed on its own (brightness, contrast, negative).
• Neighbourhood (filter) operations: new value from the pixel and its neighbours using a mask (convolution): blurring, sharpening, edge detection.
• Geometric operations: rotate, scale, crop.

Example: smoothing with a 3 x 3 averaging mask

A pixel with value 200 surrounded by neighbours 10, 20, 30, 40, 50, 60, 70, 80 (noise spike at the centre) becomes (10+20+30+40+200+50+60+70+80)/9 = 560/9 = about 62. The spike is reduced, so noise is removed but edges blur slightly.

Another example: photo of a faded old document is improved by histogram equalisation to enhance contrast, then thresholded to black and white for OCR.