The Pocket Handbook of Image Processing Algorithms in C (1993) [pdf] | Hacker News Reader

The Pocket Handbook of Image Processing Algorithms in C (1993) [pdf]

adaptiveart.eecs.umich.edu

note that this book has some significant bugs (I remember specifically a faulty quadratic equation that I was stuck on), but I don't think you use this book for the code, which is pretty outdated. This book is useful as a reminder that things like photoshop are driven by algorithms, many of which are not rocket science but rather, fit in your pocket.

One thing I particularly like about this book is that the author used photographs of his child that look super unhappy. After applying filters to them, she looks even more unhappy. It gives the book a dark and foreboding feeling.

Anyway, I always show this book to my students (along with processes like saving a raw image and opening it in audacity as a sound file, the custom tool in photoshop) as a means of demystifying image processing.

Nice looking book. Handy to keep in your back pocket in case you're walking down the street and see someone staring at a laptop and yelling in anguish, 'Please, can't someone help me implement a proper noise algorithm!? This is an emergency!'

It is unclear if this is legal to redistribute in this fashion.

The code style is making me nervous. Why use ++x over x++ when incrementing variables?

When there's a difference I find that ++x is usually what I need so I prefer to use it by default and x++ only when I need it.

20+ year C/C++ programmer here. I always use ++x unless the postfix version (x++) provides more succinct code.

The postfix operator can incur a temporary variable to be created by the compiler and thus make code slower.

Reference: http://www.idinews.com/peeves/prefixIncr.html

Also, item #6 from "More Effective C++": http://bin-login.name/ftp/pub/docs/programming_languages/cpp...

Do note that modern compilers can detect when the return value of the ++ operator is unused, and skip the extra steps when unneeded. (I've verified the resulting assembler myself from GCC 3.X.) That said, I still use the prefix notation because I can't come up with a good reason not to.

> Do note that modern compilers can detect when the return value of the ++ operator is unused, and skip the extra steps when unneeded.

Is there an equivalent instruction in assembly for postfix increment/decrement?

Neither one really has a direct "equivalent" in assembly.

Prefix is going to look something like:

  inc eax # increment the value
  mov eax,ecx # copy the return value to its destination

Postfix is going to look something like:

  mov eax,ebx # copy original value
  inc eax # increment value
  mov ebx,ecx # copy original value to final destination

The actual code you'd get would vary a lot between compilers, architectures, and even just different contexts within the same compiler-architecture combination.

Thank you, dammit I really should learn assembly.

On the flipside, using pre-increment introduces a dependency, which means the processor can't do Out-of-order execution on instructions using that value...

Published in 1993

Thanks. Added.

so patents on these can not longer be valid!

Reply on news.ycombinator.com