Optical imaging delivers absolute, non-contact, and high-dynamic-range measurement of thermal expansion. However, to achieve high accuracy, various factors should be accounted within the image analysis, including: image spatial sampling, lens aberrations, brightness nonuniformity and object edge deformations. Approach based on the object contour reconstruction is presented. Measurement procedure consists of two stages. Firstly, object edge contours corresponding to different temperatures are estimated. This is done by the novel contour-retrieving image reconstruction, capable of optical and spatial sampling superresolution as well as compensation of brightness nonuniformity. Secondly, the reference retrieved contour is reconstructed to fit retrieved contours for other temperatures, considering a linear expansion model. With the nonlinear algorithms for contour retrieval and reconstruction, small sub-resolution random edge distortions are detected and filtered out, increasing the measurement accuracy. Second improvement of the proposed approach is an opportunity to validate the measurement, given by its fully reconstructive nature.