Infrared Image Conversion from Grayscale to Temperature Using Linear Regression

A bstract This paper presents an original algorithm that aims to convert the grayscale levels of infrared images into a map of temperatures. Two methods were tried: one method which uses only the extreme temperatures provided by the legend of the infrared images; and the second one that takes into account all intermediate temperature levels with their respective grayscale references from the legend. The results obtained were compared with the actual temperature values recorded by the FLIR camera used. The linear regression conversion method was found to produce the smallest error, with a value of Mean Square Error of 0.0237, an improvement of more than 50% compared to the other method.