Abstract: A new clutter metric named C_TSSM(Clutter Metric based on Target Structure Similarity Measure) is developed for electro-optical clutter estimation. With the target structure feature space established, the background image vectors are projected into that space. By measuring the normalized length of the projections, a quantitative estimation of the structure similarity between the background and the target is obtained. C_TSSM has two advantages over those available. First, by emphasizing the spatial arrangement rather than the gray distribution of the pixels, the impact of environment variation on the stability of clutter evaluation is effectively bated. Second, with similarity measurement based on the target structure feature space and no restriction on the background constitution, C_TSSM is applicable for clutter evaluation of different kinds of scene. Experimental results show that C_TSSM can estimate the structure similarity and predict the variation of similarity with distance well.

Key words: structure similarity, clutter metric, automatic target recognition (ATR)