Multifractal Analysis of Thermal Images of Electronic Devices in Different Colour Profiles

Abstract

Nowadays, there are many methods for analysing images. Among them, a multifractal approach is able to characterise images in terms of the complexity of patterns they contain. However, multifractal methods are very sensitive to small changes in colour images. This study presents a preliminary study for the accurate characterisation of electronic devices in terms of heat dissipation using multifractal analysis. Specifically, a novel methodology using multifractal approach is proposed to characterise images in different colour formats: Grey-Scale, Red-Green-Blue (RGB), Cyan-Magenta-Yellow-Key (CMYK), Hue-Saturation-Value (HSV) and Hue-Saturation-Intensity (HSI), and the results are compared. For this purpose, thermal images of a Pi3 Raspbian Desktop board are analysed with the aim of finding out the results of the multifractal analysis on different areas of the board. Multifractal analysis is carried out through the box-counting method and the method of moments, since they allow to extract the main multifractal parameters, such as, the generalised dimensions function, D(q) and the degree of multifractality, ∆D(q). The results obtained show that both scaling properties and multifractal parameters vary according to the colour format, i.e., the same image exhibits different multifractal properties depending on the colour format.