Reduction of Video Data to the Form Typical for Measurements of the Research Object by an Ideal Sensor Based on the Eigenbasis of the Interpretation Model

The applications of the eigenbasis of the measurement interpretation model to the problem of reducing measured video data to a form typical for the measurements of the research object by an ideal measuring sensor are studied. In this case, video data are reduced as they are received and not after the recording is finished, and a researcher can stop the measurement at any time. Two methods are considered to estimate the research object image that could be obtained by an ideal measuring sensor, with an error smaller than the linear reduction error, but with a nonrandom distortion determined by the subspace of values of the research object features of interest that do not influence the result. In the first method, the research object image is estimated by selecting the components in the eigenbasis of the measurement interpretation model so that the noise component variance does not exceed a specified value and in the second method, by selecting the components of the linear reduction result by testing the hypothesis of equality of components to zero, and replacing by zero the components for which the hypothesis is not rejected. In some cases, the second method results in stronger noise suppression with less distortion.