Josh On Design

The idea of computer vision has always fascinated me. The ability to get from a plain image to an understanding of it's contents seems magical. Though I understand a bit of the underlying math, to build my own computer vision system would take years of study. Fortunately, this book and an open source library come to the rescue.

Practical Computer Vision with SimpleCV, published by O'Reilly, is a wonderful book; with an emphasis on Practical. It explains a bit about how computer vision works then dives right in to building things using the open source computer vision library: SimpleCV. SimpleCV is a Python library (linking to an underlying C implementation, I think), and quite easy to use. Since the book examples are all in Python anyone with basic programming experience (Java, C#, JavaScript, Ruby) should have no trouble with the example code.

The first half of the book starts with an explanation of basic computer vision concepts then jumps right into build a simple time lapse photography app with a few lines of Python code. Next comes image manipulations such as cropping, color reduction, simple object detection, and histograms. This is enough to create a blue screen effect and parking detector.

The second half contains the real meat of the book: detecting features. SimpleCV can pick out different shapes, filter by colors, look for faces, and even scan barcodes. One of the examples looks at a table of change to calculate the monetary value using coin size. The final chapter covers some advanced techniques like optical flow and key point matching.

While I like the book overall I do have a few nits. First, I really wish it was printed in color. Several chapters have images which can't be easily distinguished when printed in black and white. Second, I wish it was longer. While the book does covert almost every feature of SimpleCV, I'd love to read some larger example apps that combine multiple techniques. All that said, the book was still a good read and informative. It will stay on my shelf for future imaging projects.