An algorithm correcting for the fraction of scattered events in SPECT and planar imaging is proposed. The algorithm utilizes a pixel-based multi-channel analyser for data acquisition. The method was designed to operate on a local level by three subtraction steps: (a) Subtracting a modified Klein-Nishina single scatter distribution, pixel by pixel, from the events obtained experimentally in the upper half of the photo-peak window. (b) Subtracting a mirrored distribution of the unscattered events hence obtained from that of total events in the lower half of the window, thus giving the scatter distribution in this part of the window. (c) Subtracting the sum of the scatter distributions in both window halves from the corresponding sum of total events in order to obtain the unscattered photons within the photo-peak window. The accuracy of the method was validated experimentally, using a new rCBF phantom allowing for imaging in matter corresponding to soft tissue and approximately in air, respectively. After correction for photon scattering and attenuation, the regional difference in SPECT values in soft tissue equivalent matter and in low-density matter (simulating air) was only 1.5 +/- 7.2% (mean +/- 1 SD), thus indicating a high accuracy of the correction method. Provided that an accurate and stable pixel peak-alignment routine is available, the method can be applied using a minimum of three windows.