With the continued popularity of the meta-universe and the advancement of virtual reality (VR) technology, users are more eager to get a perfect immersive experience. People enter the meta-universe mainly through near-eye displays. In contrast, the advantages of silicon-based organic light-emitting diode (OLED) microdisplays such as high resolution, small size and light weight can greatly improve immersion experience. At the same time, digitally driven microdisplays have the advantages of low cost, low power consumption, high refresh rate and high contrast. However, current device bandwidths cannot support the amount of data that increases with resolution and refresh rate. In order to ensure the subjective and objective quality of the image while greatly reducing the amount of data transmission, efficient data compression technology is the key to the application.

In the digitally driven subfield scanning method of silicon microdisplay, the total display time of an image frame is often divided into several different integer or fractional subfields, and the application of just noticeable difference (JND) model to the data compression of fractional subfields has become the main direction of research. The traditional JND model is mainly inspired by the working mechanism of the human vision system, and uses the relevant characteristics of human vision to extract image features through design formulas, so as to construct the model. At present, the distortion introduced by JND model established by subjective experiment is usually additive noise such as white noise. The flat region with low image gradient is more affected by it, and the distortion is more perceptible. The distortion in the edge region with large gradient is difficult to detect. However, multiplicative noise occupies a larger proportion in actual coding, and the block effect has a greater influence on the edge region than the flat region. Therefore, we further consider the masking effect of stereovision, and design masking characteristics experiments of foveal and block effects. Meanwhile, this paper presents a cruciate block (CB) compression algorithm based on variable number of bit-planes, which adopts different number of bit-plane compression for different number of subfields. The gray level of the original image is modified by the CB-FJND model, that is, the pixel gray value is adjusted upward or downward, and the difference should meet the minimum perceptible threshold determined by the CB-FJND model. In order to improve the subjective and objective quality of the image after data compression, a cruciate block compression algorithm based on row and column alternating encoding is proposed in this paper. The row and column encoding of the binary image is adopted successively, and the subfield weights are divided according to certain strategies.

It can be seen from the 16th bit-plane distribution of the image corrected by the CB-FJND model and the change of run length according to the row run length encoding (Fig. 7 and Table 3) that, the numbers of consecutive zeros and ones increase significantly after model intervention, which could significantly improve the compression effect of subsequent run encoding. The higher 15 bit-planes in the bit-plane image present the outline of the original image, while the lower bit-planes corresponding to the fractional subfield present the details of the original image. The lower the bit plane, the more the noise and the less the details. The larger the first plane weight of the fractional subfield is, the higher the image quality is. The more dispersive the weight distribution is, the lower the image quality is. Therefore, the fractional subfield weights can be divided according to the optimal weight distribution strategy under the multi-subfield configuration, which can ensure the compression rate and improve the objective quality of the image (Tables 5 and 6). In the aspect of image coding, the CB compression algorithm using cruciate encoding method can greatly improve the subjective quality of image while ensuring the compression rate of image, and eliminate the phenomenon of horizontal and vertical stripes which may exist in the image compressed with other algorithms (Fig. 8, Fig. 11, and Table 7). Therefore, the cruciate block compression algorithm based on CB-FJND can provide better compression effect.

In this paper, based on the existing JND model of stereoscopic image, the influence factors of block effect are added, and the CB-FJND model is established to calculate the visual redundancy of the image more accurately. In order to meet the need of dynamic false contour (DFC) and gamma correction, multiple subfield bit-planes are introduced to compress the input data, and the optimal weight distribution strategy of the fractional subfield is proposed. According to the existing compression methods, a cruciate block compression algorithm is proposed. The hardware feasibility of the algorithm is verified on field programmable gate array (FPGA) platform. The experimental results show that the multiple bit-plane cruciate block compression algorithm based on CB-FJND model can greatly reduce the stereoscopic image transmission data in virtual reality to less than 40% and solve the distortion caused by excessive image data compression, ensuring the image quality, and providing a solution for increasing the user's visual experience and equipment bandwidth limitation in virtual reality.