python matplotlib plot hist2d with normalised masked numpy array

Its because of the cmin, it doesnt go well with normed=True. Removing cmin (or set it at 0) will make it work. If you do need to filter you can consider using numpy's 2d histogram function and masking the output afterwards.

a = np.random.randn(1000)
b = np.random.randn(1000)

a_ma = np.ma.masked_where(a > 0, a)
b_ma = np.ma.masked_where(b < 0, b)

bins = np.arange(-3,3.25,0.25)

fig, ax = plt.subplots(1,3, figsize=(10,3), subplot_kw={'aspect': 1})

hist, xbins, ybins, im = ax[0].hist2d(a_ma,b_ma, bins=bins, normed=True)

hist, xbins, ybins = np.histogram2d(a_ma,b_ma, bins=bins, normed=True)
extent = [xbins.min(),xbins.max(),ybins.min(),ybins.max()]

im = ax[1].imshow(hist.T, interpolation='none', origin='lower', extent=extent)
im = ax[2].imshow(np.ma.masked_where(hist == 0, hist).T, interpolation='none', origin='lower', extent=extent)

ax[0].set_title('mpl')
ax[1].set_title('numpy')
ax[2].set_title('numpy masked')