The temperature of a minute object is typically measured by an infrared thermograph, a precise and expensive device. This study presents an image–temperature model constructed by particle swarm optimization algorithm to estimate the temperature according to the color image data on object’s surface. The demonstrated case was a palladium alloy microprobe used in wafer testing. Two series of conduction experiments were implemented to imitate the probing condition of the microprobe in wafer testing. Collecting the images captured from a digital video camera and the temperatures measured by the infrared thermograph, the correlations between the color image data and their corresponding temperatures were investigated. An image–temperature model is further constructed by the particle swarm optimization algorithm based on the image components of the RGB and HSV models of the examined pixel. The model developed provides valuable references for making out the electrothermal effect of the palladium alloy microprobe used in wafer testing.