Both in medical research and clinical settings, regional hemoglobin concentrations ([Hb]) in the microcirculation of biological tissues are highly sought. Diffuse reflectance spectroscopy has been proven to be a favorable method by which to detect regional [Hb]. This paper introduces a new algorithm to retrieve [Hb] information from diffuse reflectance spectra. The proposed algorithm utilizes the natural logarithmic operation and the differential wavelet transform to effectively quench the scattering effects, and then employs the concept of isosbestic wavelength in the transformed spectra to reduce the effects of hemoglobin oxygenation. As a result, the intensity at the defined isosbestic wavelength of the transformed spectra is a good indicator of [Hb] estimation. The algorithm was derived and validated using theoretical spectra produced by Monte Carlo simulation of photon migration. Its accuracy was further evaluated using liquid tissue phantoms, and its clinical utility with an in vivo clinical study of brain tumors. The results demonstrate the applicability of the algorithm for real-time [Hb] estimations from diffuse reflectance spectra, acquired by means of a fiber-optic spectroscopy system.