Improved binary HOG algorithm and possible applications in car detection
Mariana-Eugenia Ilas · 2017
In this paper we investigate how the HOG algorithm can be successfully modified to work on binary information. We detail the area and the performance evolution when the standard HOG algorithm is adjusted to work on binary images, and introduce an improved binary HOG algorithm. Compared to the HOG working on grayscale images, the improved binary HOG has an extremely small implementation area. Its performance is practically identical to that of the original HOG working on binary images, but reduced compared to that working on grayscale images. Given its extremely small area, the new algorithm can be acceptable for many applications, or it can be used in more complex systems, as part of a preliminary detection stage.