Micro-display is an important interface connecting the real world and the metaverse world. Compared with other types of micro-displays, organic light emitting diode (OLED) on silicon micro-displays have the advantages of high resolution, high integration, low power consumption, small size, and light weight, and they have become the preferred choice of near-eye display devices. Compared with traditional OLEDs-on-silicon driven by analog signal, OLED-on-silicon micro-display driven by digital signal has obvious advantages in ultra-high-definition display. Driven by users’ demand for immersive experience of virtual reality (VR) and other near-eye display devices, near-eye displays are developing towards high resolution and high frame rate. However, this brings the problem of excessive video data transmission to the micro-display system. In order to give users a higher definition and smoother near-eye display experience, it is urgent to propose an image compression algorithm for micro-displays to solve the problem of excessive video data transmission.

Since the lower 4 bit-planes of the 19-bit-plane image reflect the details of the image and are difficult to compress, just-noticeable difference (JND) theory is introduced. In recent years, scientific research has shown that the distribution of visual acuity is asymmetric in the whole fovea range, and the human visual system has horizontal-vertical anisotropy (HVA) and vertical-meridian asymmetry (VMA). To make the JND model more consistent with the characteristics of human vision, this paper carries out psychological experiments on the luminance, contrast and foveated masking characteristics of human vision, and constructs an asymmetric elliptical foveated JND (AE-FJND) model based on the experimental results. Combined with the JND model, a corresponding bit-plane image compression algorithm is proposed. The data of the lower 4 bit-planes are processed within the JND threshold range, and then all-bit-plane images are compressed. The algorithm can compress the image without affecting the subjective perception of the human eyes. The algorithm is compared with the previously proposed compression algorithms in the following aspects: the subjective feeling of the compressed image is evaluated by the definition of the enlarged image in the center of the image, the quality of the compressed image is assessed by peak signal-to-noise ratio (PSNR), fovea PSNR (FPSNR), structural similarity (SSIM) and other evaluation indicators, and the compression performance of the algorithm is rated by compression rate.

The accuracy of the proposed AE-FJND model is verified by experiments. Subjective experiments showed that compared with the other two types of JND models, the AE-FJND model got a higher subjective score. For the same subjective score, the AE-FJND model can calculate more visual redundancy (Table 3). Objective experiments showed that the AE-FJND model had less noise distribution in the foveated area, and the amount of noise injected into the upper part of the image was greater than that in the lower part, which is consistent with the visual characteristics of human HVA and VMA (Fig.9). It can be seen from the enlarged image of the center area after noise pollution that the AE-FJND model has a high definition of the center area after noise pollution, and the result is similar to the original picture (Fig.10). In addition, the compression effect of the proposed image compression algorithm is verified. This algorithm can solve the problem that the lower 4 bit-planes cannot be compressed (Fig.12 and Table 4). The image compressed by this algorithm can maintain the same definition as the original image at the center of the image (Fig.13). Compared with a comparable compression algorithm, this algorithm can calculate more visual redundancy without affecting the subjective perception of vision, and the compressed image quality is better. In addition, the compression rate is lower than that of the comparable algorithm, which can reach 39.573%, indicating that the compression performance is better (Table 5). Compared with another compression algorithm, the proposed algorithm has higher PSNR and SSIM, and lower compression rate, which shows that this algorithm can not only ensure image quality, but also compress the image to a greater extent (Table 6).

In order to solve the problem of excessive video data transmission in high resolution and high frame rate micro-displays, this paper proposes an algorithm to compress the video data by bit-plane according to the scanning mode of digitally driven OLED-on-silicon micro-displays. Since the lower 4 bit-planes of the 19-bit-plane image reflect more details of the image and are difficult to be further compressed, the JND theory is introduced, and the two visual characteristics of the human eye, namely, HVA and VMA, are considered. An AE-FJND model is proposed, which is more consistent with the visual redundancy characteristics of the human eyes. Based on this model, a corresponding bit-plane image compression algorithm is proposed, which performs JND processing on the data of the lower 4 bit-planes, and then performs white-black-block-skip (WBBS) coding compression on different bit-planes respectively. According to the compression algorithm, the corresponding OLED-on-silicon micro-display controller is designed, and the OLED-on-silicon micro-display is successfully driven on the field programmable gate array (FPGA) platform. The bit-plane image compression algorithm based on this model can compress the image to a large extent without affecting the subjective feeling of the human eye. The average image compression rate can reach 39.573%, providing a relatively preferred solution to the problem of excessive data transmission faced by VR devices in the metaverse world.