Measuring Focus Quality in Vector Valued Images for Shape from Focus

Muhammad Tariq Mahmood, Usman Ali

Abstract: In shape from focus (SFF) methods, the focus measure (FM) operator plays a key role in determining the ultimate shape of the object. Usually, vector-valued (color) images are converted into grayscale images before applying the FM operator. This conversion saves the computations; however, it affects the accuracy of focus values which deteriorates the depth map. This paper proposes an effective FM operator to find the relative degree of focus for the vector-valued pixels in the image sequence. In the first step, vector-valued pixels are transformed into scalar-valued pixels by computing the scaled norm of the resultant vector differences. The scaling factor is computed through various features based on vector operations including dot product, cross product, projections, and vectors distances. Then differential kernels with gap are applied on the scalar image to compute the focus values. Experiments conducted using synthetic and real image sequences reveal that the proposed method is effective in providing better quality 3D shapes of the objects.