A classification of label-free super-resolution imaging mechanisms is given based on the nonlinear reduction of the point-spread function (PSF), near-field scanning, image magnification and gain, structured and sparse illumination, and information approaches. We argue that the super-resolution capability of contact microspheres stems from an image magnification effect taking place in close proximity to the object with contributions of its optical near-fields. We discuss several conditions for quantifying the super-resolution in a label-free microscopy: i) use of standalone objects or long-period arrays as opposed to subwavelength periodic structures, ii) use a convolution with two-dimensional PSF for calculating images, and iii) avoidance of coherent imaging which can lead to dramatic artifacts. We demonstrate a resolution of ~λ/7 for imaging nanoplasmonic structures and propose a combination of microspherical nanoscopy with nanoplasmonic illumination for imaging biomedical samples. We applied these techniques for imaging actin protein filaments and yeast cells and observed a resolution advantage over standard microscopy.