Image data collection, processing, storage, and their application in smartphone food analysis

A coherent set of image data collection, processing, and persistence techniques are developed in this thesis as building blocks for the solution. A spectral nonuniform illumination correction and an illumination matching algorithm are proposed to minimise any effect due to nonuniformly distributed illumination and varied illumination colour with considering variations in image formation in different cameras.

The thesis introduces a novel colour alignment technique that works with non-standard corresponding colour points and only needs a minimum of two corresponding colour points during colour matching. A high-performance camera response model that represents camera responses with only a single latent variable is proposed for more accurate and rapid camera calibration. A practical smartphone modularised colourimetric reader for on-site image analytical data collection is developed for data acquisition.