Text this: Motion Objects Segmentation and Shadow Suppressing without Background Learning