Published on August 2nd, 2012 | by Mithil Bhatia0
Halide: A New Programming Language for Image Processing and Computational Photography Software
After taking a picture, the overhead involved in displaying it on the screen is very complex and requires a large amount of processing operations. Researchers at the Massachusetts Institute of Technology’s (MIT) Computer Science and Artificial Intelligence Laboratory have constructed a new Programming Language called Halide that stands for two things – first, it simplifies the process of writing Image Processing and Computational Photography Software and second, it’s designed to economize the use of mobile battery and processors while processing Images.
For a digital camera, transforming an Image onto screen to make it appear just like how human eye sees is not an easy task. There are certain processes that have to do a lot with capturing light, reading pixels to deduce the colors, Plus-Constant-Correction process so that the result is as similar to what human eye catches, explains GigaOM. And of course, more megapixel a camera has, more intense processes and more complex and lengthy algorithms are involved.
To rewrite the algorithms in this language, researchers have been able to accelerate image processing up to three times and significantly with smaller code. In one case, Halide algorithm was actually pretty much longer than conventional MATLAB, but the Image Processing was 70x faster:
It is definitely an important step in as the web is becoming more and more visual. Halide is therefore an important programming language designed to make it easier to write short and clear image processing code on modern machines. Halide’s front end is an embedding in C++ and Hardware targets include x86-64/SSE, ARM v7/NEON, and CUDA.