This paper proposes dynamic threshold control algorithm by Co-evolutionary implementation base on a multidirectional gradient edge detection predictor (MGEDP) template. First, the image cut into four equal parts, these parts will be calculate simultaneously the error values by MGEDP predictor template, in the four sub images using, dynamic threshold evolution; Use these feedback values, to get the error image, and to calculate the threshold values by collaborative computing OTSU threshold, and so on, to classify the edges of error image, The result was convinced that this model could not only accelerated the image processing by Co-evolutionary implementation but get the more details, clearer edges, and better visual sense image.