We address how to recover a blurred image with nonuniform blur kernel when the camera moves with a constant velocity in front of a static scene. We modeled the blur kernel as a series of parameterized projective transform matrices and the blurred image as an integration of the distorted clear image sequences. By formulating the problem as a convex optimization algorithm, this problem using gradient descent algorithm can be solved. In particular, we show how to incorporate the camera’s constant movements into the projection model. By doing so, the previous nonblind nonuniform projective motion path model is changed into a blind nonuniform model. The effectiveness of our method is demonstrated by conducting experimental comparisons with other popular image deblur algorithms.