Image reconstruction techniques and target detection

Image reconstruction from projections has been extensively studied in radioastronomy and medical imaging. The same techniques can be applied to multiple target detection tasks such as radar or sonar signal processing. However, in medical imaging and radioastronomy, the images to be reconstructed are generally "compact" and the ratio between the required image size and resolution is small compared to that of the target detection problem where the images are sparse and fine resolution is required. Therefore a larger number of basis functions will be necessary to discretize the general image. Thus medical image reconstruction techniques applied directly to target detection can result in excessive memory requirements, computational time and required number of measurements. With modifications that consider the sparseness and positiveness of the images, reconstruction techniques have a potentially valuable application in the multiple target detection problem. In this paper we propose two modifications of the medical image reconstruction technique for the multi-target detection problem. One consists of pre-processing the data in order to reduce the total image to a smaller set of regions likely to contain targets. The second consists of dividing the image into pixels much larger than the expected size of a target and estimating the total target intensity for each pixel.