Wavelet Superresolution and Quasi-superresolution in Robot Vision

This article presents results of image enhancement for robot vision. Results are obtained by superresolution techniques in wavelet domain. The calculated motion of wavelet coefficients in stationary image can enhance image quality. It can be used as interpolation information for increasing resolution. This is basis for wavelet quasi-superresolution, where one low-resolution image is interpolated to high-resolution grid. Motion field of wavelet coefficients can be possibly used to predict position of image features in the next frame in multiframe approach too. Interpolation of wavelet coefficients of the first generation in combination with motion field produces weight functions. The weight functions are intuitively introduced the second generation wavelets. If this is combined with the morphology operations, the result is the third generation wavelets on the first generation settings.