HA-QL (Hidden Asynchronous Quick-Link) is a hidden modulation mode among Screen-2-Camera operating modes those are being standardized at IEEE 802.15.7m.
The previous post explains the principle to generate a hidden code on the screen using an additional intensity modulator after the bit-intensity mapper. However, the modulation of screen intensity dramatically degrades the communication performance.
In this post, an alternative solution for embedding data into the screen is introduced by applying a two-dimensional Wavelet Transform. I call this scheme is WHA-QL (Wavelet-based HA-QL)
2D Wavelet Transform
The filter-bank implementation is a good choice to implement OFDM system instead of using FFT. In our screen modulation, we do not need to design such complex carrier generation mechanism as in OFDM system, and Wavelet Transform is a good thing to start with.
The typical 2D-Inverse Discrete Wavelet Transform (2D-IDWT) reconstructs a multi-carrier modulated image from wavelet coefficients feeding on four subbands (Low-Low, Low-High, High-Low, and High-High). The LL, LH, HL, and HH coefficients contribute the approximation frequency, horizontal frequency, vertical frequency, and diagonal frequency in the space-domain to the output image respectively. In combined, four simultaneous sub-bands can be modulated to generate a coded Tx. The level of IDWT greater than one to divide a greater number of sub-bands can be an option.
Figure 1 shows experimental results for an exemplary Wavelet-MCM Screen-camera system. In this example, all of four bands can be used for modulating the intensity of screen cells. But we do NOT use all of the frequency bands! We use only high subbands for carrying data because the low-frequency band is to ensure that it is imperceptible when we embed data into an image (or a screen display).
Apply DWT for Hidden Code Generation
Overall Architecture
Figure 2 shows the Wavelet-based HA-QL system diagram.
In the system illustrated in Figure 2, we intend to embed data into an image displaying onto the screen in a manner that the change of image quality is not recognizable by our eyes. And herein, the Wavelet Transform is helpful. Two-dimensional DWT (2D-DWT) is applied with four subchannels available. The Low-Low coefficients carry processed image data, while the other three subchannels located at higher spatial-frequency domain with little image information shall be replaced by our modulated data. The replacement of higher-spatial-frequency bands will not be noticed by human eyes if the amount of data embedding into an image is reasonable.
Implementation Guide
Firstly, The Discrete Wavelet Transform (DWT) is applied to the original image to extract its lowest-frequency band. Undeniably, this lowest band contains most of the image information. After that, higher-frequency bands are used to carry data while the lowest-frequency band of the image is maintained. Our practical example in Figure 3 shows that the high frequency of the image carries the image details, and usually, our eyes cannot recognize the missing of that details.
Conclusion
In summary, we introduced the spatial-frequency modulation for hiding data into the screen image without being noticeable to our eyes. Wavelet Transform is suggested for modulating data in the spatial-frequency domain.
In practice, I’d like to combine the intensity modulation (see the previous post) and the spatial-frequency modulation (in this post) for embedding data in a hidden manner. This dedicate combination can guarantee the maintenance of the resulting image quality while providing a reliable wireless link over the screen.
Optical Camera Communication 