The topic of this project is to develop efficient algorithms for robust image description and for recovering a clear image from its degraded version. By means of moment theory, we demonstrate the original image reconstruction from its blurry copy with the convolution model. The first phase of the project involves an analytical representation of motion blur in moment domain which leads to estimation of its parameters. This approach paves the way to derive a new method for image deblurring. The second phase consists of the description of the blur model in polar coordinates instead of Cartesian ones by defining the concept of polar convolution. This can be another solution of deblurring using any radial set of moments. The final phase gives a new interpretation of blur model based on the multiplicative convolution. This concept can describe the camera motion blur along the optical axis. Easy and stable image deconvolution in moment domain is the main advantage of this model.