At present, under the multimedia environment, the accuracy of the text information extraction of video images is poor. Moreover, the extraction process is complicated. In this paper, a text information extraction algorithm for video image based on time adaptive model is proposed. Using Sobel edge detection operator, text edge of the image is extracted. Using OSTU algorithm. the adaptive threshold is obtained, to make global binary processing smooth. Through selective masking smoothing, some irrelevant background is then filtered out. By morphological method, the text area is then merged. Finally, using the connected domain, the text area is extracted, to get the text block, and obtain the text information of the video image. On this basis, according to the color feature of text in video images, through user interaction, the learning process of color online machine based on the generated time and with adaptive model can be obtained to detect the same caption in the video sequence, so as to realize the text information extraction algorithm for video image in multimedia environment. The experimental results show that the proposed algorithm can accurately extract the text information, and the speed of extraction is faster.