Time-correspondence differential ghost imaging

Experimental data with digital masks and a theoretical analysis are presented for an imaging scheme that we call time-correspondence differential ghost imaging (TCDGI). It is shown that by conditional averaging of the information from the reference detector but with the negative signals inverted, the quality of the reconstructed images is in general superior to all other ghost imaging (GI) methods to date. The advantages of both differential GI and time-correspondence GI are combined, plus less data manipulation and shorter computation time are required to obtain equivalent quality images under the same conditions. This TCDGI method offers a general approach applicable to all GI techniques, especially when objects with continuous gray tones are involved.

Figure 1

Schematic diagram of the experimental setup.

Figure 2

Top row: (a) digital mask 1, (b) DGI image, (c) GI image, and (d) CI positive minus negative image ${\langle I_R\rangle }_{+}-{\langle I_R\rangle }_{-}$, all from $140000$ frames. Middle row: (e) positive TCDGI image ${\langle I_R\rangle }_{+}^{\mathrm{diff}}$, (f) negative TCDGI image ${\langle I_R\rangle }_{-}^{\mathrm{diff}}$, (g) positive TCDGI image minus reference signal average ${\langle \delta I_R\rangle }_{+}^{\mathrm{diff}}$, and (h) negative TCDGI image minus reference signal average ${\langle \delta I_R\rangle }_{-}^{\mathrm{diff}}$, all from $69600$ frames. Bottom row: TCDGI images obtained by ${\langle I_R\rangle }_{k^{+}}^{\mathrm{diff}}-{\langle I_R\rangle }_{-k^{-}}^{\mathrm{diff}}$ with different thresholds. The numbers of frames averaged are (i) $k_3$, $6000\times 2$; (j) $k_2$, $21000\times 2$; (k) $k_1$, $31500\times 2$; and (l) $k_0=0$, $69600\times 2$.

Figure 3

The SNR vs number of reference frames.

Figure 4

The SNRs of DGI and TCDGI of different thresholds against the number of frames.

Figure 5

(a) Digital mask of Lena. Top row: DGI images from (b) $140000$, (c) 71 000, and (d) 8000 frames. Bottom row: TCDGI images from (e) $69000\times 2$, (f) $35400\times 2$, and (g) $3800\times 2$ frames for $k=1$, 2, and 3, respectively.