TV corner-logo is widely used in current video program, and becomes one of the hot spots in information extraction and analysis. TV corner-logo detection and segmentation algorithm is important for information extraction. In this paper, we presented an adaptive threshold segmentation algorithm, based on the combined time-averaged edge detection and saliency detection, to effectively separate and extract the TV corner-logo from the video sequence. First, we applied Canny operator to detect edges and then calculate the weighted average edge of ten frames, so as to get the time-averaged edge image. Then we combine the time-averaged edge image with the saliency map to get a more accurate segmentation. Finally, we used the adaptive threshold segmentation algorithm to separate the corner-logo. Experimental results show that, this method can effectively detect and separate the corner-logo from the background.