An image processing system processes images such that a proportion of area of regions of interest within the image can be increased at the expense of regions of lesser interest. First, regions of interest, such as portions of the image including participants, are determined. Then compressible background regions are determined and compressed. This results in the proportion of the area of the regions of interest to increase. After the image is stored or transmitted, the regions of interest can be seen more clearly both because they are larger and because any loss of detail caused by down-sampling or lossy image compression needed to limit the amount of image information is reduced due to the smaller image size. The process also preserves more of the relative spatial relationship between various regions of interest than prior methods.