Looks like both images you posted have both pulses and non-pulse data. What sort of classification would you like for each? To detect pulses you can run the histogram through a median filter and subtract it from the original. Then you can threshold it to find pulses taller than a certain height from their neighbors:
diffSignal = histCounts - medfilt1(histCounts);
pulseLocations = diffSignal > 100; % Or whatever number you want.
