Binary Adjacent Set Occurrence for a Simple Texture Analysis in Binary Image

Abstract

Texture analysis is one of the common methods that can be used in image classification or automated identification tasks. With the rising of IoT and microcontroller based devices which come with limited computational capabilities, choosing a simple but efficient and effective image analysis method is one of the prominent key factors in the implementation. This paper provides an alternative of simple texture analysis descriptor using binary-adjacent statistical approach. The descriptor intended to approximate coarseness or spatial distribution in the binary image using the scale-invariant feature. By scaling, it gives benefits in analyzing samples with different dimension or aspect as long as it comes from a similar region ratio. The aim is to reduce analysis complexity and memory requirement while maintaining its usability and portability. In the similarity and discrimination test, it is still able to represent better result at low scale-level compared to frequency filters of the Fourier transform method.