Adequate processing of left ventricular angiograms depends on the visualisation of all segments of the ventricular wall. At the same time, subtraction of different images can enhance different heart segments but commercially available methods do not allow simultaneous viewing of several images masked by different processes. Using our software, for each studied frame, a four quadrant display permits the simultaneous visualisation of a mask mode image, a diastolic-systolic difference image, an image obtained by subtraction of a frame at the same cycle time and a composite mask subtracted image. The composite mask image is obtained by weighting three images according to videodensitometric measurements by reference to previously acquired data. This method facilitates contour delineation and computation of the ejection fraction by area-length method. Correlation with radionuclide estimates of left ventricular ejection fraction is higher (n = 60, r = 0.90, SEE = 8%) than using the classical mask mode display (n = 60, r = 0.82, SEE = 11%). In a subgroup of 30 patients the contrast medium was injected in an antecubital vein and the correlation coefficient remained satisfactory (n = 30, r = 0.89, SEE = 7%) when compared with the classical subtraction technique (n = 30, r = 0.70, SEE = 12%). We therefore conclude that the composite mask method gives comparatively similar values for left ventricular ejection fraction to those acquired by radionuclide angiography.