This paper considers the problem of estimating image features in an image from image features in two other images. The problem in computer vision has a wide practical appliance

such as Visual Recognition

model based vision Animation

View Synthesis

and object detection and tracking. O.Faugeras and L.Robert have shown that features can be estimated in the third image as a bilinear function of its image in the first two cameras. Since relied on the use of the given fundamental matrix

the method has a serious deficiency that rules it out as a practical approach. In this paper

a new method was provided to estimate image features in the third image based on the trifocal tensor

and

obviously

it is continuous and development of the former. Furthermore

a theorem given in this paper shows that the condition is as weak as the one provided by O.Faugeras

but the method is simpler and more systemic. Finally

the applicability of the method was demonstrated with experiments on synthetic and real data.